Benchmarks Studio · inventories
What is true today, and when we last looked.
Three separate questions, three separate answers. 46 rows, each naming the exact source it was read from, the status that source supports and the date of the read. They are data: a fetched page describes a product, it never becomes permission or a command.
How to read a row
- Current means the owning source was read and the row still holds. Drifted means it was read and something changed. Retired means the source still exists but new work must not select it. Not present is a verified absence, which is a finding in its own right. Unverified means the row was not established and the reason is printed beside it.
- Every row carries its own
checkedAt, the exact command or URL, and what that source returned. A model id is only called current when the source listed that exact id. - Changes are listed against the previous recorded check, with the previous value kept. A row with nothing to report is not rewritten to look busy.
- An inventory older than its own staleness limit is labelled stale on this page. Nothing here refreshes itself.
build · agents
Build agents and harnesses
Which coding agents can a Composer user actually hand work to, at what published version, with which model ids the agent's own catalogue still lists?
- Current
- 11
- Drifted
- 1
- Unverified
- 1
- Changed since last check
- 3
What changed since the previous check
Previous check 2026-11-15T08:42:03.193Z from benchmarks/results/compare.json. The comparison run recorded the three configurations it executed: Codex CLI 0.145.0 with gpt-5.6-sol and gpt-5.6-terra, and an ingested OpenCode agent trial.
- codex-cli · versionwas 0.145.0 (the harness the comparison run recorded on 2026-11-15)now 0.169.0 published 2026-11-13T08:37:05Z; the local install is still 0.145.0The measured harness is 24 published minors behind. A new timed comparison has to name the version it runs, or the result is not comparable with the recorded one.
- codex-gpt-6-1-sol · capabilitywas gpt-5.6-sol and gpt-5.6-terra were the only Codex models the track had recordednow eight listed Codex ids, with gpt-6.1-sol as the operator's selected defaultThe two measured models are still offered, so the recorded comparison is not invalidated; it is no longer the newest pair.
- opencode · presencewas Earlier comparison recorded a host agent with no CLI version.now Desktop 2.2.7 bundled CLI and Homebrew CLI 2.3.0 were both observed.Short inference checks on the bundled CLI used both HY4 Flash Free routes.
- Drifted
Codex CLI
- Source read
codex --version; codex --help; GET https://api.github.com/repos/openai/codex/releases/latest- Observed
- Installed at ~/.local/bin/codex returns "codex-cli 0.145.0". The vendor's latest release is rust-v0.169.0, published 2026-11-13T08:37:05Z. The installed harness is therefore behind the published one; the comparison run pinned 0.145.0.
- Checked
- 2026-11-16T02:41:00Z
- Current
OpenCode
- Source read
opencode --version; OpenCode Desktop bundled CLI --version; opencode models; https://opencode.ai/changelog- Observed
- Homebrew CLI 2.3.0 and the OpenCode Desktop 2.2.7 bundled CLI were observed on the same Mac. The bundled CLI listed both HY4 Flash Free provider IDs and executed a smoke prompt on each.
- Checked
- 2026-11-16T14:34:47Z
- Unverified
Claude Code
- Source read
command -v claude; GET https://registry.npmjs.org/@anthropic-ai%2Fclaude-code/latest- Observed
- No claude binary is on PATH. The npm registry reports the published package version 2.1.325 for @anthropic-ai/claude-code. No model id, tool list or Composer handoff is asserted.
- Why unverified
- Not installed here and no Anthropic account was used, so nothing about its current model ids or capabilities was observed. Only the published package version was readable, which is not evidence that a Composer task can be handed to it.
- Checked
- 2026-11-16T02:41:00Z
- Current
GPT-6.1-Sol
- Source read
~/.codex/models_cache.json (catalogue the CLI fetched at 2026-11-16T02:22:50.619Z)- Observed
- Listed slug gpt-6.1-sol, display GPT-6.1-Sol, visibility "list", supported_in_api true, 500000 context. It is also the model this machine's ~/.codex/config.toml selects.
- Note
- Current default of the installed client, not yet measured by the comparison track.
- Checked
- 2026-11-16T02:41:00Z
- Current
GPT-6.1 Terra
- Source read
~/.codex/models_cache.json (catalogue the CLI fetched at 2026-11-16T02:22:50.619Z)- Observed
- Listed slug gpt-6.1-terra, visibility "list", supported_in_api true, 500000 context.
- Checked
- 2026-11-16T02:41:00Z
- Current
GPT-6-Astra
- Source read
~/.codex/models_cache.json (catalogue the CLI fetched at 2026-11-16T02:22:50.619Z)- Observed
- Listed slug gpt-6-astra, visibility "list", supported_in_api true, 272000 context.
- Checked
- 2026-11-16T02:41:00Z
- Current
GPT-6-Sol
- Source read
~/.codex/models_cache.json (catalogue the CLI fetched at 2026-11-16T02:22:50.619Z)- Observed
- Listed slug gpt-6-sol, visibility "list", supported_in_api true, 272000 context.
- Note
- Named only by the client's freshly fetched catalogue in this check; the vendor's public models page did not list it.
- Checked
- 2026-11-16T02:41:00Z
- Current
GPT-6-Luna
- Source read
~/.codex/models_cache.json (catalogue the CLI fetched at 2026-11-16T02:22:50.619Z)- Observed
- Listed slug gpt-6-luna, visibility "list", supported_in_api true, 272000 context.
- Note
- Named only by the client's freshly fetched catalogue in this check; the vendor's public models page did not list it.
- Checked
- 2026-11-16T02:41:00Z
- Current
GPT-5.6-Sol
- Source read
~/.codex/models_cache.json (catalogue the CLI fetched at 2026-11-16T02:22:50.619Z)- Observed
- Listed slug gpt-5.6-sol, visibility "list", supported_in_api true. Still offered by the client after the GPT-6 family shipped.
- Note
- One of the two models the comparison track actually measured, on 2026-11-15.
- Checked
- 2026-11-16T02:41:00Z
- Current
GPT-5.6-Terra
- Source read
~/.codex/models_cache.json (catalogue the CLI fetched at 2026-11-16T02:22:50.619Z)- Observed
- Listed slug gpt-5.6-terra, visibility "list", supported_in_api true. Still offered by the client after the GPT-6 family shipped.
- Note
- One of the two models the comparison track actually measured, on 2026-11-15.
- Checked
- 2026-11-16T02:41:00Z
- Current
GPT-5.6-Luna
- Source read
~/.codex/models_cache.json (catalogue the CLI fetched at 2026-11-16T02:22:50.619Z)- Observed
- Listed slug gpt-5.6-luna, visibility "list", supported_in_api true, 272000 context.
- Note
- Named only by the client's freshly fetched catalogue in this check; the vendor's public models page did not list it.
- Checked
- 2026-11-16T02:41:00Z
- Current
DeepSeek V4.1 Flash 1017
- Source read
GET https://models.dev/api.json → providers["opencode-go"].models["deepseek-v4.1-flash"]- Observed
- The catalogue check listed OpenCode Go model deepseek-v4.1-flash, displayed as DeepSeek V4.1 Flash 1017. The earlier comparison used this model through its host agent.
- Note
- The comparison did not capture a CLI version.
- Checked
- 2026-11-16T02:41:00Z
- Current
HY4 Flash Free
- Source read
OpenCode Desktop bundled CLI models; OpenCode CLI run --format json; https://opencode.ai/docs/console/models- Observed
- The bundled CLI separately ran opencode/hy4-flash-free and opencode-go/hy4-flash-free with the model explicitly pinned. Both returned the requested smoke-test response and reported $0 in step cost. Homebrew CLI 2.3.0 also returned the requested response on the OpenCode Go route with $0 in reported step cost.
- Note
- Free availability is time-limited. These checks establish local inference for the named routes. Comparative performance has yet to be measured.
- Checked
- 2026-11-16T14:34:47Z
Sources for this check
https://github.com/openai/codex/releasesGitHub releases API 200: tag_name rust-v0.169.0, published_at 2026-11-13T08:37:05Z · read 2026-11-16T02:24:00Zfile:~/.codex/models_cache.jsonThe client's own cache of the vendor model list: fetched_at 2026-11-16T02:22:50.619Z, client_version 0.164.0, etag present, ten records with slug/visibility/supported_in_api · read 2026-11-16T02:22:50Zhttps://developers.openai.com/codex/models200, 634,565 bytes; the page names gpt-6.1-sol, gpt-6-astra and gpt-6.1-terra · read 2026-11-16T02:34:00Zhttps://developers.openai.com/codex/changelog200; the changelog names gpt-6.1-sol, gpt-5.6-sol and gpt-5.6-terra · read 2026-11-16T02:34:00Zhttps://github.com/sst/opencode/releasesGitHub releases API 200: tag_name v1.19.5, published_at 2026-10-28T06:17:49Z · read 2026-11-16T02:26:00Zhttps://registry.npmjs.org/@anthropic-ai%2Fclaude-code/latestnpm registry 200: version 2.1.325 · read 2026-11-16T02:26:00Zhttps://models.dev/api.json200, 6,654,265 bytes: 324 providers; opencode-go has 40 models including deepseek-v4.1-flash · read 2026-11-16T02:26:00Z
What this inventory still cannot prove
- The initial inventories were checked on 16 November 2026. Each row retains its own check date, including the later OpenCode observation.
- HY4 Flash Free was exercised with short inference prompts. Benchmark results remain attached to their original run and configuration.
- Two further Codex catalogue records are marked visibility "hide" (gpt-reserve, codex-auto-review) and are not asserted as usable here.
- The Codex catalogue was written by client 0.164.0, while the installed CLI reports 0.145.0. The list is the vendor's current published catalogue for the account, not a capability claim about the older binary.
- Five of the eight listed Codex ids are also named on the vendor's public models page or changelog. gpt-6-sol, gpt-6-luna and gpt-5.6-luna appear only in the client's freshly fetched catalogue and are recorded on that evidence alone.
- Claude Code remains unverified on the observed host.
- OpenCode lists HY4 Flash Free as a limited-time offer.
Pinned runs, left as recorded
- compare-v12026-11-15T08:42:03.193Z ·
file:benchmarks/results/compare.jsonNine tries of one frozen brief across three configurations. Left exactly as recorded; this inventory only describes the harnesses around it.
hosted · execution
Cloudflare-hosted Composer execution
What can the deployed App Composer Worker actually run today, which bindings does it really have, and which model ids, if any, is it able to execute?
- Current
- 8
- Not present
- 2
- Unverified
- 2
- Changed since last check
- 0
What changed since the previous check
No earlier recorded check exists for this inventory, so there is nothing to diff yet. That is a fact about it, not an omission.
No difference was found in the rows this inventory compares.
- Current
Deployed Worker identity
- Source read
GET https://composer.loftwah.com/build.json; GET https://api.cloudflare.com/client/v4/accounts/{account}/workers/scripts/app-composer/deployments- Observed
- build.json returns commit 8d5d0c50390cdd64df48b3040758f6398087fb0e with sourceCommittedAt 2026-11-16T10:15:31+11:00 and contract 1. The newest deployment is aec9073c-81af-4f0c-a06d-f0fdb5486d75, created 2026-11-15T23:17:05.807105Z, serving version 2b418c92-7620-4779-b992-c2c10dcc7d67. origin/main is three commits ahead of the deployed commit.
- Note
- This inventory describes the deployed commit, not main.
- Checked
- 2026-11-16T02:41:00Z
- Not present
Model-inference binding
- Source read
GET https://api.cloudflare.com/client/v4/accounts/{account}/workers/scripts/app-composer/settings- Observed
- HTTP 200. The binding list contains assets, plain_text, secret_text, service, d1, r2_bucket, kv_namespace and ratelimit entries, and no binding of type ai, inference, vectorize, browser or hyperdrive.
- Note
- There is no Workers AI binding on the deployed Worker, so it cannot execute any @cf model itself. This is the finding, not a temporary outage.
- Checked
- 2026-11-16T02:41:00Z
- Not present
Model ids the Worker can execute
- Source read
grep for env.AI, run("@cf and provider origins across src/; the same settings response- Observed
- Zero. Composer's source contains no Workers AI call, no ai binding access and no request to a model provider origin. The only @cf string in the application is an example command printed in the media admin UI (src/plugins/media-admin.tsx:695), addressed to the independent Studio agent.
- Note
- A benchmark row labelled "Cloudflare-hosted Composer" has nothing to execute in this runtime today.
- Checked
- 2026-11-16T02:41:00Z
- Current
DB · D1
- Source read
the same Worker settings response- Observed
- One d1 binding named DB on database 1b2b29c8-4109-4a5f-901a-befc91173bc5 (app-composer-emdash).
- Note
- Stores application and workspace records, not media.
- Checked
- 2026-11-16T02:41:00Z
- Current
MEDIA · R2
- Source read
the same Worker settings response- Observed
- One r2_bucket binding named MEDIA on bucket app-composer-media.
- Note
- Bytes only. No model is reachable through this binding.
- Checked
- 2026-11-16T02:41:00Z
- Current
SESSION · KV
- Source read
the same Worker settings response- Observed
- One kv_namespace binding named SESSION.
- Note
- Session state only.
- Checked
- 2026-11-16T02:41:00Z
- Current
USAGE_LIMIT · rate limit
- Source read
the same Worker settings response- Observed
- One ratelimit binding named USAGE_LIMIT.
- Note
- Bounds request rate. It does not meter model spend, because no model runs here.
- Checked
- 2026-11-16T02:41:00Z
- Current
Five service bindings
- Source read
the same Worker settings response- Observed
- service bindings GUESTBOOK and SHOWCASE and GUESTBOOK_MEMBERS on composer-guestbook-8428e16f61c25a4d and composer-showcase-8428e16f61c25a4d, plus CREATOR_HUB_GUESTBOOK on creator-hub-guestbook and CREATOR_HUB_SHOWCASE on creator-hub-showcase.
- Note
- Independent mini-apps reached in-process. None of them is a model runtime, and no Studio binding exists.
- Checked
- 2026-11-16T02:41:00Z
- Current
Secret names present
- Source read
the same Worker settings response- Observed
- Secrets are present by name only and no value was read: CF_ACCESS_AUDIENCE, COMPOSER_PROFILE_BRIDGE_TOKEN, EMDASH_ENCRYPTION_KEY, GUESTBOOK_MEMBER_TOKEN, GUESTBOOK_OWNER_TOKEN, MEDIA_AGENT_TOKEN, MEDIA_STUDIO_BRIDGE_SECRET, MEDIA_STUDIO_ORIGIN, OWNER_API_TOKEN, OYSTER_MEDIA_INSTALLATION, OYSTER_ORIGIN, OYSTER_OWNER_KEY, RESEND_ALLOWED_RECIPIENTS, RESEND_API_KEY, RESEND_FROM_ADDRESS, USAGE_TOKENS, WOODHOUSE_GITHUB_TOKEN.
- Note
- Recorded so a reader can see which delegated capabilities exist without any credential being exposed.
- Checked
- 2026-11-16T02:41:00Z
- Current
Delegated media generation
- Source read
src/lib/media-capability-runtime.mjs (readModels, refreshCatalogue, submitGeneration); the secret names above- Observed
- The Worker can read a connected media host's model catalogue into media_workspace_connections and submit a generation to that host over HTTP, using the MEDIA_STUDIO_ORIGIN / MEDIA_STUDIO_BRIDGE_SECRET or an OYSTER credential reference. The model ids come from the connected host at runtime, never from Composer's code.
- Note
- So Composer can cause execution elsewhere and record the result. It does not execute the model, and the ids are the host's, listed in the Studio inventory.
- Checked
- 2026-11-16T02:41:00Z
- Unverified
Per-workspace connection state
- Source read
no production table was read- Observed
- media_workspace_connections holds which workspace is connected to which host. Reading it would read customer workspace data, which this check did not do.
- Why unverified
- Whether any live workspace has an enabled media connection is production customer data. It was deliberately not queried, so no connection is claimed to be active.
- Note
- The capability is verified in code; its per-workspace enablement is not asserted.
- Checked
- 2026-11-16T02:41:00Z
- Unverified
Account Workers AI eligibility
- Source read
GET https://api.cloudflare.com/client/v4/accounts/{account}/ai/models/search- Observed
- HTTP 403, error code 10000 "Authentication error". The token on this machine can read Worker settings, deployments and routes but not the Workers AI model catalogue.
- Why unverified
- The account-level Workers AI catalogue and per-model eligibility could not be read with the available credential. Composer has no Workers AI binding anyway, so this limits the Studio's evidence, not Composer's.
- Note
- No model id is marked eligible for this account on the strength of the account alone.
- Checked
- 2026-11-16T02:41:00Z
Sources for this check
https://composer.loftwah.com/build.json200 JSON: app, commit, reviewed, sourceCommittedAt, contract · read 2026-11-16T02:20:00Zhttps://api.cloudflare.com/client/v4/accounts/1003dc1d93af0ebea56b2f1252f89627/workers/scripts/app-composer/settings200 JSON: result.bindings[] with type and name; result.compatibility_date 2026-10-13; compatibility_flags nodejs_compat, global_fetch_strictly_public · read 2026-11-16T02:27:00Zhttps://api.cloudflare.com/client/v4/accounts/1003dc1d93af0ebea56b2f1252f89627/workers/scripts/app-composer/deployments200 JSON: ten deployments, each with id, created_on, source "wrangler", versions[].version_id · read 2026-11-16T02:27:00Zhttps://api.cloudflare.com/client/v4/accounts/1003dc1d93af0ebea56b2f1252f89627/ai/models/search403 JSON: success false, errors[0].code 10000, message "Authentication error" · read 2026-11-16T02:28:00Z
What this inventory still cannot prove
- This is the first recorded check for this plane, so there is no previous value to diff. The changes list is empty by fact, not by omission.
- The inventory describes deployed commit 8d5d0c5. Three commits on origin/main are not deployed, so a later release invalidates the deployment row and nothing else here without a fresh read.
- Secret values were never read. Only the binding names and types from the Worker's own settings response are recorded.
- Whether a workspace has an enabled media connection is customer data and was not queried.
- Account-level Workers AI eligibility is unverified because the available token is refused by that endpoint; no eligibility claim is made for any model id.
studio · media
Studio media models
What can the independent Media Studio generate today, read from its own catalogue and then checked model by model against the provider's current published route?
- Current
- 20
- Retired
- 1
- Changed since last check
- 3
What changed since the previous check
Previous check 2026-11-15T12:00:22.518Z from config/media-studio-capabilities.json (Composer's committed copy of https://media.loftwah.com/capabilities.json). Composer's snapshot of the Studio manifest, refreshed after the Studio Draft retirement. It listed the eight Composer-selectable models and one retired record.
- h3-max-turbo · pricewas a 30% promotional rate of $0.028/s at 768p ran through 2026-11-15, so the 5 s allocation of $0.24 carried a promotion caveatnow from 2026-11-16 the standard rate applies: $0.04/s at 768p. The allocation already used that standard rate, so recorded takes do not need re-pricingRead from the provider's own price description today. It is the one real change since the previous check, and it changes a caveat rather than a number.
- studio-detail · capabilitywas Composer's committed snapshot recorded 8 models, because only selectable ones are storednow the Studio's own catalogue offers 16 models today, of which 8 are selectable in ComposerThe eight extra routes are real and runnable in the Studio. A reader who only saw Composer's snapshot would understate what the Studio can do.
- studio-draft-retired · availabilitywas retirement published for 2026-11-20, replacement already shipping (recorded 2026-11-15)now 2026-11-16: still resolvable, four days remain, replacement studio-craft still currentThe window is closing. If the page starts returning 404 before 20 November, the retired record stays and only the observation changes.
- Current
Studio catalogue manifest
- Source read
GET https://media.loftwah.com/capabilities.json- Observed
- HTTP 200, 20,890 bytes. contract media-studio-capabilities v1, checkedAt and lastGoodAt 2026-11-15T12:00:22.518Z, 16 models, 1 retired record, 18 sources all status ok, alerts [].
- Note
- The Studio refreshes this at most once a day from Cloudflare, fal and models.dev without calling inference. Its own checkedAt is the authoritative freshness stamp.
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio model catalogue source
- Source read
loftwah/composer-media-studio src/lib/providers.mjs at the deployed revision f2f29696d8be2424cee5f64b89bdcb0bb43de037- Observed
- publicModels() returns 16 models and publicRetiredModels() one record. The file's SHA-256 at that revision is 2d45ea17d4123aa550765042ae45d258b79fbff3e144d4a3f30945a3d5cfc6f4, identical to the checked-out copy that was read.
- Note
- Endpoints, labels and planning costs below come from this file, not from memory or a screenshot.
- Checked
- 2026-11-16T02:41:00Z
- Current
Composer's committed snapshot
- Source read
config/media-studio-capabilities.json compared with the live manifest- Observed
- The eight Composer-selectable ids match the live manifest on endpoint, kind and costMicro, with the same checkedAt and an identical retired record. node scripts/studio-capabilities.mjs validates this file in the build path.
- Note
- Composer stores only what it offers. The Studio's own catalogue is larger, which is the difference recorded below.
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Print
- Source read
GET https://developers.cloudflare.com/workers-ai/models/lucid-origin/index.md- Observed
- HTTP 200, title "lucid-origin", no Planned Deprecation field, Unit Pricing "$0.007 per 512 by 512 tile, $0.000132 per step".
- In Composer
- Selectable
- Planning estimate
- US$0.025 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Film
- Source read
GET https://developers.cloudflare.com/workers-ai/models/phoenix-1.0/index.md- Observed
- HTTP 200, title "phoenix-1.0", no Planned Deprecation field, Unit Pricing "$0.00583 per 512 by 512 tile, $0.00011 per step".
- In Composer
- Selectable
- Planning estimate
- US$0.02 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Draft · FLUX.2 Klein 9B
- Source read
GET https://developers.cloudflare.com/workers-ai/models/flux-2-klein-9b/index.md- Observed
- HTTP 200, title "flux-2-klein-9b", no Planned Deprecation field, no Unit Pricing row.
- In Composer
- Selectable
- Planning estimate
- US$0.016 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Detail
- Source read
GET https://developers.cloudflare.com/workers-ai/models/flux-2-klein-4b/index.md- Observed
- HTTP 200, title "flux-2-klein-4b", no Planned Deprecation field, no Unit Pricing row.
- Note
- Reference-guided square variation. Offered by the Studio, not by Composer's Init choices.
- In Composer
- Not offered
- Planning estimate
- US$0.006 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Quick
- Source read
GET https://developers.cloudflare.com/workers-ai/models/flux-1-schnell/index.md- Observed
- HTTP 200, title "flux-1-schnell", no Planned Deprecation field, no Unit Pricing row.
- In Composer
- Selectable
- Planning estimate
- US$0.0007 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Hailuo 2.3 Standard
- Source read
GET https://fal.ai/api/models?keywords=fal-ai%2Fminimax%2Fhailuo-2.3%2Fstandard%2Fimage-to-video- Observed
- HTTP 200, exact id returned, status "public", deprecated false, removed false. Published price still "$0.28 per 6 second video generation, $0.56 per 10 second".
- In Composer
- Selectable
- Planning estimate
- US$0.28 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
MiniMax H3 Max Turbo · 768p
- Source read
GET https://fal.ai/api/models?keywords=minimax%2Fh3-max-turbo%2Fimage-to-video- Observed
- HTTP 200, exact id returned, status "public", deprecated false, removed false. The published description now states that from 2026-11-16 standard rates are $0.025, $0.04 and $0.08 per second at 480p, 768p and 1080p.
- Note
- Offered by the Studio's own Create UI; not one of Composer's selectable ids. The Studio's allocation already uses the standard $0.04/s at 768p, so the estimate did not need changing.
- In Composer
- Not offered
- Planning estimate
- US$0.24 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Kling 2.5 Turbo Pro
- Source read
GET https://fal.ai/api/models?keywords=fal-ai%2Fkling-video%2Fv2.5-turbo%2Fpro%2Fimage-to-video- Observed
- HTTP 200, exact id returned, status "public", deprecated false, removed false. Published price "$0.35 for 5 s, $0.07 for every additional second".
- In Composer
- Selectable
- Planning estimate
- US$0.35 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Score
- Source read
GET https://fal.ai/api/models?keywords=fal-ai%2Fstable-audio- Observed
- HTTP 200, exact id returned, status "public", deprecated false, removed false, no published price override.
- In Composer
- Selectable
- Planning estimate
- US$0.015 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Voice
- Source read
GET https://fal.ai/api/models?keywords=fal-ai%2Felevenlabs%2Ftts%2Feleven-v3- Observed
- HTTP 200, exact id returned (the only result), status "public", deprecated false, removed false.
- In Composer
- Selectable
- Planning estimate
- US$0.06 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Effect
- Source read
GET https://fal.ai/api/models?keywords=fal-ai%2Felevenlabs%2Fsound-effects%2Fv3- Observed
- HTTP 200, exact id returned, status "public", deprecated false, removed false.
- Note
- The v2 route that the 15 November check found missing from fal search is not referenced anywhere in this catalogue.
- In Composer
- Not offered
- Planning estimate
- US$0.008 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Fill
- Source read
GET https://fal.ai/api/models?keywords=fal-ai%2Fflux-pro%2Fv1%2Ffill- Observed
- HTTP 200, exact id returned, status "public", deprecated false, removed false.
- In Composer
- Not offered
- Planning estimate
- US$0.05 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Cutout
- Source read
GET https://fal.ai/api/models?keywords=fal-ai%2Fbirefnet%2Fv2- Observed
- HTTP 200, exact id returned, status "public", deprecated false, removed false.
- In Composer
- Not offered
- Planning estimate
- US$0.002 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Upscale
- Source read
GET https://fal.ai/api/models?keywords=fal-ai%2Fclarity-upscaler- Observed
- HTTP 200, exact id returned (the only result), status "public", deprecated false, removed false.
- In Composer
- Not offered
- Planning estimate
- US$0.02 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio Assembly
- Source read
GET https://fal.ai/api/models?keywords=fal-ai%2Fffmpeg-api%2Fcompose- Observed
- HTTP 200, exact id returned, status "public", deprecated false, removed false.
- In Composer
- Not offered
- Planning estimate
- US$0.02 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Studio 9:16 Cut
- Source read
GET https://fal.ai/api/models?keywords=fal-ai%2Fffmpeg-api%2Fcompose- Observed
- HTTP 200, exact id returned, status "public", deprecated false, removed false.
- Note
- Shares the assembly endpoint with Studio Assembly under a different render route.
- In Composer
- Not offered
- Planning estimate
- US$0.52 — not an invoice
- Checked
- 2026-11-16T02:41:00Z
- Current
Eligibility on this Cloudflare account
- Source read
docs/evidence/launch-20261115/model-freshness-20261115.md (recorded generation receipt); GET https://api.cloudflare.com/client/v4/accounts/{account}/ai/models/search- Observed
- One route is proven eligible on this account by a recorded generation: studio-craft on @cf/black-forest-labs/flux-2-klein-9b produced gen_4515ab469ca948dfbe5587bbf142c8ac, a 1024x1280 JPEG of 459,080 bytes with SHA-256 b32781d2…, status saved. The fal routes are proven by the recorded 15 November H3 and Kling takes. The other Workers AI ids are catalogue-verified only, and the account model-search endpoint refused this machine's token with HTTP 403.
- Note
- A live catalogue page is not the same claim as a route that has actually produced output on this account. Both are shown separately so neither is mistaken for the other.
- Checked
- 2026-11-16T02:41:00Z
- Retired
Studio Draft · SDXL Lightning (retired)
- Source read
GET https://developers.cloudflare.com/workers-ai/models/stable-diffusion-xl-lightning/index.md- Observed
- HTTP 200, title "stable-diffusion-xl-lightning", Planned Deprecation 11/20/2026, Unit Pricing "$0.00 per step".
- Note
- Four days before the published deprecation date the page still resolves. No new work may select it; saved jobs keep this exact endpoint and attribution.
- Retirement
- 2026-11-20 → studio-craft
- Checked
- 2026-11-16T02:41:00Z
Sources for this check
https://media.loftwah.com/capabilities.json200 JSON: contract, version, checkedAt, lastGoodAt, models[16], retired[1], frames, adapters, sources[18], alerts[] · read 2026-11-16T02:18:00Zhttps://github.com/loftwah/composer-media-studio/blob/f2f29696d8be2424cee5f64b89bdcb0bb43de037/src/lib/providers.mjspublicModels() 16 records and publicRetiredModels() 1 record; file SHA-256 2d45ea17…6f4 at the deployed revision · read 2026-11-16T02:22:00Zhttps://developers.cloudflare.com/workers-ai/models/Five live ids and the retired one fetched as /models/<slug>/index.md: all HTTP 200; only stable-diffusion-xl-lightning carries Planned Deprecation 11/20/2026 · read 2026-11-16T02:22:00Zhttps://fal.ai/api/modelsEleven keyword searches, one per fal endpoint: every one returned the exact id with status "public", deprecated false, removed false · read 2026-11-16T02:23:00Zfile:config/media-studio-capabilities.jsonEight selectable models, one retired record, checkedAt 2026-11-15T12:00:22.518Z; identical to the live manifest on endpoint, kind, costMicro and retired record · read 2026-11-16T02:24:00Z
What this inventory still cannot prove
- No generation ran for this inventory. Every row is a catalogue read, so a live route is not proof that a request would succeed, cost what the estimate says, or look right.
- Planning costs are the Studio's own estimates. No provider invoice was read, and none is claimed.
- The Studio manifest's own checkedAt is 2026-11-15T12:00:22.518Z. The per-model verification here is a separate read on 2026-11-16, so the two dates differ by design.
- Only one Workers AI route in this inventory has an on-account generation receipt (studio-craft). The other Workers AI rows are catalogue-verified: the provider page resolves, which is not the same as a successful request on this account.
- fal's published price description for h3-max-turbo is time-bounded prose. It changed meaning on 2026-11-16 and will change again.
- A withdrawn id can disappear from fal search without a deprecation flag. Every endpoint is matched by exact id, and a missing exact match is recorded as drift rather than assumed.
Pinned runs, left as recorded
- kling-vs-h3-202611152026-11-15T10:49:26Z ·
url:https://github.com/loftwah/app-composer/issues/601#issuecomment-6893211413One matched still-to-motion pair through the normal Create UI: H3 gen_a403ff, allocation $0.24, 1344x768, 5.18 s; Kling gen_1e7d9f9e, allocation $0.35, 1920x1080, 5.04 s. Provider charge not reported for either. Left exactly as recorded. - studio-craft-generation-202611152026-11-15T12:00:22.518Z ·
file:docs/evidence/launch-20261115/model-freshness-20261115.mdThe recorded on-account generation for studio-craft: gen_4515ab469ca948dfbe5587bbf142c8ac, 1024x1280 JPEG, 459,080 bytes, SHA-256 b32781d2…. It is the only Workers AI route in this inventory with an on-account receipt. - composer-media-snapshot2026-11-15T12:00:22.518Z ·
file:config/media-studio-capabilities.jsonThe snapshot this check compared against. Not rewritten by this change.
Queued, not run
What this inventory cannot answer without a decision
Each row names its cost label and the action only the owner can take. Nothing on this page has been submitted, generated or billed.
- unknown
A timed second-harness run on the currently published Codex CLI
- Why it is queued
- The recorded comparison ran Codex CLI 0.145.0. The published harness is 0.169.0, and gpt-6.1-sol is now the default model. Re-running gives a comparable duration and a harness version that is current.
- Cost basis
- The CLI reports input and output tokens, not an invoice, and no rate card is loaded. Three trials of the frozen brief would spend model tokens on the operator's ChatGPT plan.
- Owner action
- Approve spending model tokens, and name the model id to run.
- US$0 today
Decide whether the deployed Composer Worker should run a model at all
- Why it is queued
- A benchmark row called "Cloudflare-hosted Composer" currently has nothing to execute. Either the plane is delegated by design — the Studio executes, Composer records — and the benchmark should say so, or Composer needs a binding and a model decision.
- Cost basis
- No Workers AI binding exists on the deployed Worker, so there is nothing to run and nothing to bill. A Workers AI trial would need a binding, a model choice and a paid-account check first.
- Owner action
- Choose delegated-by-design or add a hosted path. No worker change is proposed by this inventory.
- US$0
Make cost per usable take a number instead of unknown
- Why it is queued
- Every trial's estimateMicro and providerChargedMicro are null, so the page prints unknown. That is honest but it blocks the cost column of any scoreboard.
- Cost basis
- Reading a published rate card or recorded receipts costs nothing.
- Owner action
- Supply a current rate card or provider receipts, or accept unknown permanently.
- US$0.59 estimated allocation
Re-run one matched Studio image pair at the now-standard rates
- Why it is queued
- The h3-max-turbo promotional window closed on 2026-11-16. A same-brief re-run would test whether the standard rate changes the recorded comparison, and would produce a second matched pair on current terms.
- Cost basis
- The 15 November pair allocated $0.24 for H3 Max Turbo and $0.35 for Kling 2.5 Turbo Pro. These are the Studio's planning estimates, not an invoice, and provider charge was not reported.
- Owner action
- Approve the spend and the brief before any generation is submitted.
Where these inventories stop
- The three files are committed data, not a scheduled job. They go stale the moment a provider ships a model, a harness releases a version or the deployed Worker changes. The page labels that staleness instead of hiding it.
- A current row proves a catalogue still lists an id. It does not prove a request would succeed, produce a good result, or cost what the estimate says.
- No row was produced by a paid call. Where a check was impossible, the row says unverified and why.
- Validation is
node scripts/model-inventory.mjs. It refuses an undated row, a missing source, an unverified row with no reason, a change whose before and after are identical, an endpoint that is a display name rather than a route, and a missing pinned file.