fix(models): registry audit: new Together/Fireworks/Gemini/Mistral/xAI models, xai retirement repricing, bedrock grok-4.6 caching, deprecation dates - #38804
Conversation
… native audio (from #38573), anthropic sources Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
…uired for gemini models Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
…evin/1787944648-registry-audit-rolling
…ai/gemini deprecation dates Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
…evin/1787944648-registry-audit-rolling
…tries - gemini: nano-banana-pro-preview, gemma-4-26b-a4b-it, gemma-4-31b-it - mistral: 14 official aliases from api.mistral.ai/v1/models - fireworks_ai: glm-5p3, qwen3-embedding-8b - xai: grok-imagine-video, grok-imagine-video-1.5 Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
…, and unsourced entries Reprice ten more retired xAI slugs (grok-3 and grok-3-mini families, grok-4-1-fast) to the grok-4.3 rates they now bill at, with family-correct deprecation dates. Restore cache_read_input_token_cost on the Bedrock Grok 4.6 entries so implicit cache hits bill at the cache-read rate while explicit cachePoint stays unsupported. Drop the unsourced 1080p video rate and the gemini/ live native-audio entry the Gemini API 404s on. Add Groq qwen3.8-27b tool-use flags per Groq docs. Extend the xai and gemini tests to lock all of this in
|
|
Greptile SummaryThis PR updates LiteLLM’s model registry with new provider models, corrected pricing and deprecation metadata, capability adjustments, and corresponding regression coverage.
Confidence Score: 5/5The PR appears safe to merge because the previously reported unsupported xAI video registrations have been fully removed and no blocking failure remains. No blocking failure remains.
|
| Filename | Overview |
|---|---|
| model_prices_and_context_window.json | Updates the primary registry with new models, corrected pricing, capabilities, endpoints, and deprecation dates while removing the unsupported xAI video entries. |
| litellm/model_prices_and_context_window_backup.json | Mirrors the primary registry changes, including removal of both unsupported xAI video entries. |
| tests/test_litellm/integrations/test_anthropic_cache_control_hook.py | Extends coverage for Bedrock Grok 4.6 explicit-caching support and implicit cache-read pricing metadata. |
| tests/test_litellm/llms/gemini/realtime/test_gemini_realtime_transformation.py | Verifies the Vertex-only native-audio model is exposed through the realtime endpoint. |
| tests/test_litellm/llms/xai/test_xai_cost_calculator.py | Updates xAI cost-calculation coverage for the revised retired-model pricing. |
| tests/test_litellm/llms/xai/test_xai_redirected_slug_pricing.py | Expands coverage across redirected xAI slugs and checks that dated retired chat aliases are included. |
| tests/test_litellm/test_together_ai_model_metadata.py | Adds metadata and serverless-registry coverage for Together AI GLM-5.3. |
Reviews (3): Last reviewed commit: "fix(registry): add gemma 4 capability fl..." | Re-trigger Greptile
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| "litellm_provider": "xai", | ||
| "mode": "video_generation", |
There was a problem hiding this comment.
When a caller invokes the video-generation API with either newly registered xAI video model, provider dispatch cannot construct an xAI video configuration and raises ValueError: video generation is not supported for xai before sending an upstream request.
Knowledge Base Used: Provider adapters and capabilities
|
bugbot run |
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.
Autofix Details
Bugbot Autofix prepared a fix for the issue found in the latest run.
- ✅ Fixed: Gemma 4 missing capability flags
- Added supports_function_calling, supports_tool_choice, supports_response_schema, and supports_vision flags to the gemini/gemma-4-26b-a4b-it and gemini/gemma-4-31b-it entries in both registry files to match the sibling gemma-3-27b-it capability set.
Or push these changes by commenting:
@cursor push c3c4912329
Preview (c3c4912329)
diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json
--- a/litellm/model_prices_and_context_window_backup.json
+++ b/litellm/model_prices_and_context_window_backup.json
@@ -23614,7 +23614,11 @@
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
+ "supports_function_calling": true,
"supports_reasoning": true,
+ "supports_response_schema": true,
+ "supports_tool_choice": true,
+ "supports_vision": true,
"source": "https://ai.google.dev/gemini-api/docs/pricing"
},
"gemini/gemma-4-31b-it": {
@@ -23625,7 +23629,11 @@
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
+ "supports_function_calling": true,
"supports_reasoning": true,
+ "supports_response_schema": true,
+ "supports_tool_choice": true,
+ "supports_vision": true,
"source": "https://ai.google.dev/gemini-api/docs/pricing"
},
"gemini/imagen-3.0-fast-generate-001": {
diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json
--- a/model_prices_and_context_window.json
+++ b/model_prices_and_context_window.json
@@ -23614,7 +23614,11 @@
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
+ "supports_function_calling": true,
"supports_reasoning": true,
+ "supports_response_schema": true,
+ "supports_tool_choice": true,
+ "supports_vision": true,
"source": "https://ai.google.dev/gemini-api/docs/pricing"
},
"gemini/gemma-4-31b-it": {
@@ -23625,7 +23629,11 @@
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
+ "supports_function_calling": true,
"supports_reasoning": true,
+ "supports_response_schema": true,
+ "supports_tool_choice": true,
+ "supports_vision": true,
"source": "https://ai.google.dev/gemini-api/docs/pricing"
},
"gemini/imagen-3.0-fast-generate-001": {You can send follow-ups to the cloud agent here.
|
bugbot run |
There was a problem hiding this comment.
✅ Bugbot reviewed your changes and found no new issues!
Comment @cursor review or bugbot run to trigger another review on this PR
Reviewed by Cursor Bugbot for commit cfb7a26. Configure here.
a979c89
into
litellm_internal_staging
3 participants
Replaces #38693 at its head 2bd7b58: same commits and authors, moved to a
litellm_branch because CircleCI's branch filter never builds adevin/headTLDR
Problem this solves:
/v1/realtimeHow it solves it:
/v1/realtimeto that Vertex entry and sources to the Claude 5 entriesdeprecation_datevalues from official provider pagesUser Flow
Before: a developer calling GLM-5.3 through Together AI gets answers that are never charged, so their key's spend and the logs page stay at $0
"model": "together_ai/zai-org/GLM-5.3"and a short promptx-litellm-response-costheader at all andx-litellm-key-spend: 0.0After: the same call is priced at Together's published GLM-5.3 rates, so key spend and the logs page show real cost
"model": "together_ai/zai-org/GLM-5.3"and the same promptx-litellm-response-cost: 0.0003054($1.40/M input, $4.40/M output), andx-litellm-key-spendrises by that amountRelevant issues
Supports #26900 (deprecation metadata)
Linear ticket
Resolves LIT-6472
Changes in the review pass (2026-08-29, third pass)
xAI, per https://docs.x.ai/developers/migration/may-15-retirement and https://docs.x.ai/developers/pricing:
"model": "grok-4.3"and the proxy prices from the returned model. The repricing fixes the registry surfaces (/model_group/info,/model/info, the Models page, andmodel_map_informationin spend logs), which still showed the retired ratesBedrock Grok 4.6, per the AWS model card:
cache_read_input_token_cost($0.55/M us, $0.50/M global) on both entries: AWS documents implicit prompt caching with cache-read billing for this modelsupports_prompt_cachingstays false: the flag drivessupports_prompt_caching()and model info, and it stops the cache-control hook from auto-injecting breakpoints. Acache_controlblock the caller puts in the request is still forwarded as acachePoint, which Bedrock rejects with 403 for this model (see QA case 5), same as before this PR. Without the rate, implicit cache hits would bill at the full input rateGemini:
gemini/gemini-live-2.5-flash-native-audioentry an earlier commit of this PR had added (from feat(pricing): add gemini-live-2.5-flash-native-audio to model cost map #38573): the Gemini API returns 404 for it on v1beta, v1alpha, and v1, and the Live websocket rejects the model, so against the base branch this PR adds no Gemini API entry for it. The existing Vertexgemini-live-2.5-flash-native-audioentry gains/v1/realtimeinsupported_endpoints, and the realtime test asserts the model stays Vertex-onlyGroq, per https://console.groq.com/docs/model/qwen3.8-27b:
groq/qwen/qwen3.8-27b(new in this PR) carriessupports_function_calling,supports_response_schema, andsupports_tool_choice(the doc lists Tool Use, JSON Object Mode, and JSON Schema Mode)Together:
together_ai/zai-org/GLM-5.3added to the serverless coverage test listmax_output_tokens; LIT-6489 tracks fixing the sync's context-length fallback and sweeping those entriesChanges in the second pass (2026-08-29)
Together AI, per https://docs.together.ai/docs/serverless-models and https://www.together.ai/models/glm-5-3 (absorbed from #38733 with corrections):
together_ai/zai-org/GLM-5.3($1.40/M in, $4.40/M out, $0.26/M cached, 1048575 ctx). Corrected feat(models): sync together_ai model registry #38733'smax_output_tokensfrom 1048575 to 128000 (the documented output ceiling) and added the capability flags feat(models): sync together_ai model registry #38733 flagged as missing (tools, reasoning, response schema, prompt caching)together_ai/Qwen/Qwen3.8-2.4T-A95Brepriced: in 2e-06 -> 2.5e-06, out 6e-06 -> 6.25e-06, cache read 2.5e-07 -> 5e-07Gemini, per https://ai.google.dev/gemini-api/docs/pricing and
GET generativelanguage.googleapis.com/v1beta/models:gemini/nano-banana-pro-preview(API alias for Gemini 3 Pro Image; priced at the published gemini-3-pro-image rates, 131072 input limit from the models API)gemini/gemma-4-26b-a4b-itandgemini/gemma-4-31b-it(free of charge per the pricing page; 262144/32768 limits from the models API). Capability flags (vision, function calling, tool choice, response schema, system messages, reasoning) confirmed against the live Gemini API: inline image,function_declarations,functionCallingConfig.mode: ANY,responseSchema, andsystem_instructionall return 200 on both sizes, while audio input returns 400Audio input modality is not enabled for this model, matching the model card (audio is E2B/E4B/12B only)Mistral, per
GET api.mistral.ai/v1/models(the API's ownaliasesfield): 14 official alias entries cloned from their canonical registry entries:mistral-medium-3.5,mistral-vibe-cli-latest,mistral-vibe-cli-with-tools,mistral-vibe-cli-fast,mistral-code-latest,mistral-code-fim-latest,mistral-code-agent-latest,mistral-ocr-3,mistral-ocr-3-0,mistral-ocr-4,voxtral-mini-latest,voxtral-mini-realtime-2602,voxtral-mini-realtime-latest,labs-leanstral-1-5-1Fireworks AI, per https://docs.fireworks.ai/serverless/pricing and
GET api.fireworks.ai/inference/v1/models:fireworks_ai/accounts/fireworks/models/glm-5p3($1.40/M in, $4.40/M out, $0.26/M cached; 1048576 ctx from the models API, 128000 output per the GLM-5.3 model docs)fireworks_ai/accounts/fireworks/models/qwen3-embedding-8b($0.10/M input per the embeddings pricing table, 40960 ctx)Chose not to change:
grok-imagine-videoandgrok-imagine-video-1.5($0.05 and $0.08 per second of output, per https://docs.x.ai/developers/models/grok-imagine-video and https://docs.x.ai/developers/models/grok-imagine-video-1.5): an earlier commit of this PR added them, but the proxy has no xAI video adapter, so a caller picking them onPOST /v1/videosgetsvideo generation is not supported for xaiinstead of a video. Greptile flagged this on the review; the entries are dropped until an adapter existsglm-5p3-flashand thekimi-k2p6-turborouter: in the models API but no row on the pricing pageantigravity-preview-*,deep-research-*: billed at underlying model rates, no fixed per-token price to record;aqa: no published pricingChecked, nothing missing or wrong: OpenAI (models API vs registry; computer-use-preview and current snapshots present), Anthropic (models API; entries current; the three Claude 5 entries gain a
sourceURL). DeepSeek model-list API rejected the configured key, so that provider was not audited this run; Groq was checked against its public docs in the review pass. OpenRouter/Bedrock/Azure/Cohere/Perplexity/Nebius/Voyage were not reached this run; they stay on the audit backlog.Changes in the first pass (2026-08-29)
xAI May 15, 2026 retirements, per https://docs.x.ai/developers/migration/may-15-retirement and https://docs.x.ai/developers/models/grok-4.3 and https://docs.x.ai/developers/models/grok-build-0.1: retired chat slugs (grok-3, grok-4, grok-4-0709, grok-4-1-fast-, grok-4-fast-) now priced at grok-4.3 rates (1.25e-06 in, 2.5e-06 out, 2e-07 cache read, 200K tier at 2.5e-06/5e-06/4e-07); stale 128K tiers removed. grok-code-fast* aliases priced at grok-build-0.1 rates (1e-06 in, 2e-06 out, 2e-07 cache read)
Bedrock Grok 4.6 (
us.xai.grok-4.6,global.xai.grok-4.6): explicit prompt caching flag removed; Bedrock rejectscachePointbreakpoints for this model. The review pass above restores the implicit cache-read billing this initially droppedOpenAI deprecation dates, per https://platform.openai.com/docs/deprecations:
deprecation_date: 2026-12-01added to the 15 sized/qualitygpt-image-1.5-2025-12-16variants (the base snapshot entry already had it)Gemini, per https://ai.google.dev/gemini-api/docs/deprecations:
gemini/gemini-omni-flash-previewgetsdeprecation_date: 2026-09-30(replacement gemini-omni-1.1-flash). The vertex_ai copy was left alone because the Vertex deprecations page does not list itVerified as current, no change needed: Anthropic dates against https://docs.anthropic.com/en/docs/about-claude/model-deprecations, other Gemini dates, OpenAI exact-name dates
Absorbed from other PRs
Dropped as unverifiable
deprecation_date: 2026-02-28on Grok-2-era xAI models: the announcement URL it cites returns 404 and no live official xAI page states that date. Itsmode: responseschanges are already on the base branchNot absorbed (code/behavior or structural PRs that only incidentally touch the JSON): #36233, #29898, #30383, #36416, #36422
Pre-Submission checklist
Please complete all items before asking a LiteLLM maintainer to review your PR
uv run pytest tests/test_litellm/<your_test_file>.py -v. Leave the suites (make test-unit-*,make test-unit) to CI: it finishes in ~15 minutes where a laptop takes an hour or more@greptileaito re-request a review after pushing changes)Delays in PR merge?
If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).
Screenshots / Proof of Fix
Both legs boot the proxy from the named commit with
LITELLM_LOCAL_MODEL_COST_MAP=True,--num_workers 2, a fresh Postgres DB, and this config ($KEYis the master key,$PORTis the random port each leg picked):Every request is a real provider call. The grok-3-mini cases hit all three unified endpoints because the repricing applies to each of them
Before (c24f821)
/v1/chat/completions with xai/grok-3-mini
curl -sD - http://localhost:$PORT/v1/chat/completions -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "xai/grok-3-mini", "messages": [{"role": "user", "content": "Say hi in five words"}], "max_tokens": 64}'cost_in_usd_ticks(10721500 = $0.00107215) agrees. The old registry rates (3e-07 in, 5e-07 out) would give 0.0002214/v1/messages with xai/grok-3-mini
curl -sD - http://localhost:$PORT/v1/messages -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "xai/grok-3-mini", "max_tokens": 64, "messages": [{"role": "user", "content": "Say hi in five words"}]}'/v1/responses with xai/grok-3-mini
curl -sD - http://localhost:$PORT/v1/responses -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "xai/grok-3-mini", "input": "Say hi in five words", "max_output_tokens": 64}'/v1/chat/completions with together_ai/zai-org/GLM-5.3
curl -sD - http://localhost:$PORT/v1/chat/completions -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "together_ai/zai-org/GLM-5.3", "messages": [{"role": "user", "content": "Say hi in five words"}], "max_tokens": 64}'x-litellm-response-cost*header of any kind):/v1/chat/completions with bedrock-grok-46 and an explicit cache_control block
curl -sD - http://localhost:$PORT/v1/chat/completions -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "bedrock-grok-46", "messages": [{"role": "system", "content": [{"type": "text", "text": "You are terse.", "cache_control": {"type": "ephemeral"}}]}, {"role": "user", "content": "Say hi in five words"}], "max_tokens": 64}'HTTP/1.1 500withYour session has expired ... Please reauthenticate using 'aws login'instead of a Bedrock answer. The After side below shows the request-level behavior is the same with the flag off, so the missing Before capture does not change the verdict/spend/logs row for the grok-3-mini chat call
curl -sD - "http://localhost:$PORT/spend/logs?request_id=e53c192e-adb5-9dc1-b3ab-cd7058f2c83c" -H "Authorization: Bearer $KEY"model_map_informationstill shows the retired rates/model/info and /v1/chat/completions with gemini/gemma-4-31b-it and gemini/gemma-4-26b-a4b-it (no-DB proxy, 1 worker, both models from a
gemini/config entry)curl -s http://localhost:$PORT/model/info -H "Authorization: Bearer $KEY" | jq '.data[] | select(.model_name=="gemini/gemma-4-31b-it") | .model_info | {supports_vision, supports_function_calling, supports_tool_choice, supports_response_schema, supports_system_messages, supports_reasoning, max_input_tokens, input_cost_per_token}'curl -sD - http://localhost:$PORT/v1/chat/completions -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "gemini/gemma-4-31b-it", "messages": [{"role": "system", "content": "Answer in one word."}, {"role": "user", "content": [{"type": "text", "text": "What color is this image?"}, {"type": "image_url", "image_url": {"url": "data:image/png;base64,<1x1 green png>"}}]}]}'curl -s http://localhost:$PORT/v1/chat/completions -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "gemini/gemma-4-26b-a4b-it", "messages": [{"role": "user", "content": "Weather in Boston?"}], "tools": [{"type": "function", "function": {"name": "get_weather", "parameters": {"type": "object", "properties": {"city": {"type": "string"}}, "required": ["city"]}}}], "tool_choice": "required"}'After (2bd7b58)
Captured at 2bd7b58; the commits since, 68404d8 (deletes the two
xai/grok-imagine-video*registry entries) and cfb7a26 (adds capability flags to the twogemini/gemma-4-*entries), touch entries none of the cases above call/v1/chat/completions with xai/grok-3-mini
curl -sD - http://localhost:$PORT/v1/chat/completions -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "xai/grok-3-mini", "messages": [{"role": "user", "content": "Say hi in five words"}], "max_tokens": 64}'/v1/messages with xai/grok-3-mini
curl -sD - http://localhost:$PORT/v1/messages -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "xai/grok-3-mini", "max_tokens": 64, "messages": [{"role": "user", "content": "Say hi in five words"}]}'/v1/responses with xai/grok-3-mini
curl -sD - http://localhost:$PORT/v1/responses -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "xai/grok-3-mini", "input": "Say hi in five words", "max_output_tokens": 64}'/v1/chat/completions with together_ai/zai-org/GLM-5.3
curl -sD - http://localhost:$PORT/v1/chat/completions -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "together_ai/zai-org/GLM-5.3", "messages": [{"role": "user", "content": "Say hi in five words"}], "max_tokens": 64}'/v1/chat/completions with bedrock-grok-46 and an explicit cache_control block
curl -sD - http://localhost:$PORT/v1/chat/completions -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "bedrock-grok-46", "messages": [{"role": "system", "content": [{"type": "text", "text": "You are terse.", "cache_control": {"type": "ephemeral"}}]}, {"role": "user", "content": "Say hi in five words"}], "max_tokens": 64}'cache_control:curl -sD - http://localhost:$PORT/v1/chat/completions -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"model": "bedrock-grok-46", "messages": [{"role": "system", "content": "You are terse."}, {"role": "user", "content": "Say hi in five words"}], "max_tokens": 64}'/spend/logs row for the grok-3-mini chat call
curl -sD - "http://localhost:$PORT/spend/logs?request_id=05e0d200-756b-9162-a1d1-f111999aae71" -H "Authorization: Bearer $KEY"(a fresh grok-3-mini chat call on this leg, headerx-litellm-response-cost: 0.0013921500000000002)model_map_informationnow carries the grok-4.3 rates/model/info and /v1/chat/completions with gemini/gemma-4-31b-it and gemini/gemma-4-26b-a4b-it (captured at cfb7a26, the tip)
GET /model/infoas the Before legtool_choice: required, andresponse_format: json_schemacalls as the Before leg/model/info,/model_group/info, the Models page, andlitellm.supports_*now report what the Gemini API actually accepts for these two modelsPending: a screenshot of https://litellm-domain/ui/?page=logs showing the GLM-5.3 request at $0 (Before) and $0.0003 (After); the QA run had no browser
Closing notes from the run:
/spend/logs?request_id=matches the responseid, notx-litellm-call-id; unchangedLive PR risk check
Registry-only change, so the dependent paths are the cost calculator, the capability and caching flags, and the read-only registry surfaces. Live A/B at the merge base and the tip, both proxies with 2 workers and
LITELLM_LOCAL_MODEL_COST_MAP=True:Cost calculator: covered by the six cases above, real provider calls on all three unified endpoints
GET /model/infofor nine deployments (xai/grok-3-mini, xai/grok-4-1-fast, xai/grok-code-fast-1, together_ai/zai-org/GLM-5.3, groq/qwen/qwen3.8-27b, bedrock us.xai.grok-4.6, vertex gemini-live-2.5-flash-native-audio, mistral/mistral-medium-3.5, gemini/gemini-omni-flash-preview): the diff is exactly the intended changes (grok-3-mini and grok-4-1-fast at grok-4.3 rates with the 200K tier and no 128K tier, Bedrock Grok 4.6supports_prompt_cachingfalse, the three new models populated instead of null, the Vertex live entry listing/v1/realtime). grok-code-fast-1 and gemini-omni-flash-preview are identical on both sides;deprecation_dateis not exposed on this route on either sideBedrock Grok 4.6 with a caller-supplied
cache_control: 403 from Bedrock at the tip, same as before the flag change (see case 5)Legacy suites that read the registry, run locally at the tip:
tests/llm_translation/test_xai.pyandtests/llm_translation/realtime/test_xai_realtime.py(34 passed, 23 skipped),tests/litellm_utils_tests/test_utils.py -k "supports_reasoning or supports_web_search or grok"(13 passed)CircleCI finding:
.circleci/config.ymlfilters every job tomainand/litellm_.*/, so the originaldevin/head of fix(models): registry audit: new Together/Fireworks/Gemini/Mistral/xAI models, xai retirement repricing, bedrock grok-4.6 caching, deprecation dates #38693 produced a pipeline with zero workflows underrun-ci; this PR exists so the same commits get a real CircleCI runGemma 4 flags (cfb7a26):
litellm.supports_vision,supports_function_calling,supports_response_schema,supports_system_messages, andsupports_reasoningreturn True for both entries withLITELLM_LOCAL_MODEL_COST_MAP=True;tests/test_litellm/test_model_prices_schema.pyandtests/local_testing/test_get_model_info.pypass (49 tests);ci_cd/check_files_match.pypassescfb7a26 passes /live-pr-risk
Type
🐛 Bug Fix
Caveats (if any)
Medium
Low
/v1/realtime; not live-testedFinal Attestation
Link to Devin session: https://app.devin.ai/sessions/1eb47e52d5a0484985cc0e51592871d5
Open in Devin Desktop: https://app.devin.ai/desktop/session/1eb47e52d5a0484985cc0e51592871d5?variant=devin