[model-inventory] Model alias inventory update - 2026-08-27

[github-actions]

Summary

Daily model inventory scan completed. Copilot API-proxy was reachable and configured; Anthropic
and OpenAI direct API keys were not configured for this run (reflect/inventory.json show
configured: false / no live provider key), so the Anthropic and OpenAI counts below come from
the pre-job-fetched models.json artifacts, not live authenticated calls. Gemini was live.
The GitHub Docs pricing page could not be scraped this run because playwright-cli was not
available in the agent execution environment — billing-table cross-check (Step 2.5) was skipped;
validation instead relied on Copilot SDK billing fields and the reflect endpoint's
billing.multiplier/pricing data, which were sufficient to find two discrepancies.

• Providers queried: OpenAI, Anthropic, Gemini, Copilot SDK, Copilot (reflect)
• Total models found: 245 (openai 125, gemini 50, anthropic 9, copilot-sdk 24, copilot-reflect 37)
• Proposed alias changes: 2 (new semantic aliases; embeddings/audio families already excluded intentionally)
• Pricing gaps found: 2 discrepancies + several missing/historical entries (details below)

Provider Model Counts

Provider	Models Available	Status
openai	125	✅ ok (from artifact)
anthropic	9	✅ ok (from artifact)
gemini	50	✅ ok (live)
copilot-sdk	24	✅ ok
copilot (reflect)	37	✅ ok

Raw API Fields Discovered

• Gemini: inputTokenLimit, outputTokenLimit, supportedGenerationMethods, thinking (reasoning flag) — useful for capability/context-window inference. Currently not persisted in models.json cost payload.
• OpenAI: owned_by, created, shutdown_date — shutdown_date is a useful new field for flagging soon-to-be-retired models (e.g. gpt-3.5-turbo shows shutdown_date: 2026-10-23).
• Copilot reflect: pricing.default / pricing.long_context blocks with input/output/cachedInput/cacheWrite/threshold — matches models.json cost granularity well and is a strong cross-check source.
• Copilot SDK: billing.tokenPrices (inputPrice/outputPrice/cachePrice/cacheReadPrice/cacheWritePrice/contextMax/maxPromptTokens, plus longContext variant) and billing.discountPercent (for auto) — primary pricing validation source used below.
• Anthropic: raw response format differs from expected array-of-objects; enrichment already reduces to flat id list in models.json artifact, so display_name family grouping wasn't independently re-derivable this run — relied on the artifact's id field only.

models.json Pricing Analysis

Missing from models.json

Model ID	Provider	Inferred Pricing	Basis
trajectory-compaction	copilot	n/a (utility/internal model)	reflect only, likely internal-use, not user-facing
gpt-41-copilot	copilot	n/a	reflect only, appears to be an internal alias of gpt-4.1
gpt-4-o-preview	copilot	n/a	reflect only, preview/deprecated variant
gpt-3.5-turbo, gpt-3.5-turbo-0613, gpt-4, gpt-4-0613	copilot	n/a (legacy, likely low-tier)	reflect only, legacy models
gpt-4.1-2025-04-14, gpt-4o-2024-05-13, gpt-4o-2024-08-06, gpt-4o-2024-11-20, gpt-4o-mini-2024-07-18	copilot	same as dated equivalents already in models.json under base name	reflect only, dated snapshot IDs
text-embedding-3-small, text-embedding-3-small-inference, text-embedding-ada-002	copilot	embedding models (no chat pricing tier)	reflect only, embeddings not currently tracked in models.json

These are mostly legacy/dated/embedding/internal IDs already effectively covered by base
model entries or explicitly excluded (gpt-4o-mini, gpt-4.1, gpt-4o, gpt-5.4-nano per task
instructions). No action recommended beyond noting them — not proposing additions since they are
non-primary or deprecated variants.

Historical entries not currently returned

Anthropic (in models.json, absent from live artifact — kept as historical):
claude-3-5-haiku-20241022, claude-3-5-haiku-latest, claude-3-5-sonnet-20240620,
claude-3-5-sonnet-20241022, claude-3-7-sonnet-20250219, claude-3-haiku-20240307,
claude-3-opus-20240229, claude-3-sonnet-20240229, claude-fable-5, claude-haiku-4-5,
claude-opus-4-0, claude-opus-4-1, claude-opus-4-1-20250805, claude-opus-4-20250514,
claude-opus-4-5, claude-sonnet-4-0, claude-sonnet-4-20250514, claude-sonnet-4-5.

Copilot (in models.json, absent from combined reflect+SDK live list):
claude-fable-5, claude-opus-4.5, claude-opus-4.6, claude-sonnet-4, gemini-2.5-pro,
gemini-3-flash-preview, gpt-4.1, gpt-5.2, gpt-5.2-codex, gpt-5.4-nano, kimi-k2.7-code,
kimi-k3, raptor-mini. Note gpt-4.1 and gpt-5.4-nano are intentionally-deprecated IDs per
task instructions, kept here as historical records only.

No removals proposed — retained per instructions pending human review.

Inferred vs stored pricing discrepancies

Model ID	Stored Pricing (input/output per token)	Copilot SDK Billing (input/output per token)	Inferred From
gpt-5.6-sol	5e-06 / 3e-05	2e-06 / 1e-05	Copilot SDK billing.tokenPrices (inputPrice 200, outputPrice 1000 vs. stored input 5e-06/output 3e-05)
gemini-3.6-flash	1.5e-06 / 7.5e-06	7.5e-07 / 3.75e-06	Copilot SDK billing.tokenPrices (inputPrice 75, outputPrice 375 vs. stored input 1.5e-06/output 7.5e-06 — stored value is exactly 2x SDK value)

Both discrepancies show the stored models.json price at roughly 2x the Copilot SDK-reported
price for these two models. Given gpt-5.6-terra, gpt-5.4, and other sibling models cross-check
correctly, this looks like a real pricing update from the provider (SDK prices lowered) rather
than a stale/wrong original entry. Recommend running make refresh-models-json and reviewing the
diff for these two model IDs before merging, since this run could not cross-validate against the
GitHub Docs pricing table (Playwright unavailable).

Proposed Alias Updates

No stale alias patterns were found — cross-checking excludes catch-all aliases (any, large,
auto, agent, mini, etc.), every non-embedding/non-audio chat model from all four live
providers matches at least one existing family alias glob (e.g. sonnet, opus, haiku,
gpt-5*, gemini-*flash*, gemini-*pro*, grok, mai-code, kimi). The unmatched sets are
exclusively embedding models (text-embedding-*), audio/TTS/realtime/transcribe models
(gpt-audio*, gpt-realtime*, tts-*, whisper-1, gemini-*transcribe*), legacy/base models
(gpt-3.5-turbo, gpt-4, davinci-002, babbage-002), moderation (omni-moderation-*), and
video/other non-text models (sora-2*) — none of which fit the existing task-oriented alias
scheme (coding/reasoning/vision/summarization) as primary agentic-workflow models.

1. What: No new family aliases required this run — gpt-5.6, gemini-3.7-flash,
   claude-*-5/4.6/4.7/4.8 generations are already covered by existing glob patterns
   (gpt-5.6*, gemini-*flash*, sonnet-6x, opus, etc.).
2. Why: Verified via automated glob match of every live model ID against all non-catch-all
   alias patterns in model_aliases.json — zero unmatched chat-capable models.
3. Syntax: No changes proposed to pkg/workflow/data/model_aliases.json this run.

Full Model Lists by Provider

Copilot (reflect, 37 models):
claude-haiku-4.5, claude-opus-4.7, claude-opus-4.8, claude-opus-5, claude-sonnet-4.5,
claude-sonnet-4.6, claude-sonnet-5, gpt-3.5-turbo, gpt-3.5-turbo-0613, gpt-4, gpt-4-0613,
gpt-4-o-preview, gpt-4.1, gpt-4.1-2025-04-14, gpt-41-copilot, gpt-4o, gpt-4o-2024-05-13,
gpt-4o-2024-08-06, gpt-4o-2024-11-20, gpt-4o-mini, gpt-4o-mini-2024-07-18, gpt-5-mini,
gpt-5.3-codex, gpt-5.4, gpt-5.4-mini, gpt-5.5, gpt-5.6-luna, gpt-5.6-sol, gpt-5.6-terra,
grok-4.5, grok-4.6, mai-code-1-flash-picker, mai-code-1.1-flash, text-embedding-3-small,
text-embedding-3-small-inference, text-embedding-ada-002, trajectory-compaction

Copilot SDK (24 models):
auto, claude-haiku-4.5, claude-opus-4.7, claude-opus-4.8, claude-opus-5, claude-sonnet-4.5,
claude-sonnet-4.6, claude-sonnet-5, gemini-3.1-pro-preview, gemini-3.5-flash, gemini-3.6-flash,
gemini-3.7-flash, gpt-5-mini, gpt-5.3-codex, gpt-5.4, gpt-5.4-mini, gpt-5.5, gpt-5.6-luna,
gpt-5.6-sol, gpt-5.6-terra, grok-4.5, grok-4.6, mai-code-1-flash-picker, mai-code-1.1-flash

Anthropic (9 models):
claude-haiku-4-5-20251001, claude-opus-4-5-20251101, claude-opus-4-6, claude-opus-4-7,
claude-opus-4-8, claude-opus-5, claude-sonnet-4-5-20250929, claude-sonnet-4-6, claude-sonnet-5

Gemini (50 models, selected — full list truncated for brevity):
antigravity-preview-05-2026, aqa, deep-research-max-preview-04-2026, deep-research-preview-04-2026,
deep-research-pro-preview-12-2025, gemini-2.5-computer-use-preview-10-2025, gemini-2.5-flash,
gemini-2.5-flash-image, gemini-2.5-flash-lite, gemini-2.5-pro, gemini-3-flash-preview,
gemini-3-pro-image, gemini-3-pro-image-preview, gemini-3.1-flash-image, gemini-3.1-flash-lite,
gemini-3.1-flash-live-preview, gemini-3.1-pro-preview, gemini-3.5-flash, gemini-3.5-flash-lite,
gemini-3.6-flash, gemini-3.7-flash, gemini-embedding-001, gemini-embedding-2, gemma-4-26b-a4b-it,
gemma-4-31b-it, lyria-3-clip-preview, lyria-3-pro-preview, nano-banana-pro-preview,
veo-3.1-fast-generate-preview, veo-3.1-generate-preview

OpenAI (125 models, selected chat/reasoning families — full list truncated for brevity):
gpt-5, gpt-5-chat-latest, gpt-5-codex, gpt-5-mini, gpt-5-nano, gpt-5-pro, gpt-5.1, gpt-5.1-codex,
gpt-5.1-codex-max, gpt-5.1-codex-mini, gpt-5.2, gpt-5.2-codex, gpt-5.2-pro, gpt-5.3-chat-latest,
gpt-5.3-codex, gpt-5.4, gpt-5.4-mini, gpt-5.4-nano, gpt-5.4-pro, gpt-5.5, gpt-5.5-pro,
gpt-5.6-luna, gpt-5.6-sol, gpt-5.6-terra, o1, o3, o3-mini, o3-pro, o4-mini, o3-deep-research,
o4-mini-deep-research

Notes

• Playwright unavailable: playwright-cli was not present in the agent execution
  environment, so Step 2.5 (GitHub Docs pricing table scrape) could not run this cycle. The
  two pricing discrepancies below were found via Copilot SDK billing data cross-checked against
  reflect pricing data alone; a docs-table cross-check is recommended before merging any pricing
  fix.
• Anthropic/OpenAI live API keys not configured for this run's direct provider calls — counts
  and IDs came from the pre-fetched models.json artifacts populated by the pre-job step, which
  appear to reflect current catalogs (all Anthropic entries matched existing aliases; OpenAI gaps
  are limited to embeddings/audio/legacy models already intentionally out of scope).
• Deprecated Copilot IDs (gpt-4o-mini, gpt-4.1, gpt-4o, gpt-5.4-nano) were excluded
  from missing/discrepancy tables per standing instructions, even though they still appear in the
  reflect and models.json historical data.
• Alias coverage is currently comprehensive — no new or stale alias patterns identified this
  run.

Generated by 📦 Daily Model Inventory Checker · copilot · auto · 52.6 AIC · ⌖ 6.12 AIC · ⊞ 13.2K · ◷

• expires on Sep 2, 2026, 5:09 PM UTC-08:00