[sorenjuul]

Summary

Adds the text-generation models that have been added to the Abacus (RouteLLM) catalog since #4705, sourced from the live models endpoint. All entries point to existing lab metadata via base_model and override only provider-specific cost, real limit/modality deltas, and reasoning_options.

Models added

Abacus ID	Base model
route-llm-code	full inline (RouteLLM router SKU, like route-llm.toml)
route-llm-code-low	full inline (RouteLLM router SKU, like route-llm.toml)
gemini-3.8-flash	google/gemini-3.8-flash
grok-4.7	xai/grok-4.7
claude-fable-5-1	anthropic/claude-fable-5-1
claude-opus-5-5-thinking	anthropic/claude-opus-5-5
gpt-6-astra	openai/gpt-6-astra
gpt-6-sol	openai/gpt-6-sol
gpt-6-luna	openai/gpt-6-luna
muse-spark-1.3	meta/muse-spark-1.3
muse-spark-1.3-contributor	meta/muse-spark-1.3 (contributor pricing tier of the same model)
zai-org/GLM-5.3	zhipuai/glm-5.3
zai-org/GLM-5.3-Flash	zhipuai/glm-5.3-flash
deepseek-ai/DeepSeek-V4.1-Flash	deepseek/deepseek-v4.1-flash
deepseek-ai/DeepSeek-V4-Pro-0813	deepseek/deepseek-v4-pro-0813
deepseek-ai/DeepSeek-V4-Flash-0731	deepseek/deepseek-v4-flash-0731
deepseek-ai/DeepSeek-V4-Flash-Vision-Exp	deepseek/deepseek-v4-flash-vision-exp
mimo-v2.6-pro	xiaomi/mimo-v2.6-pro
moonshotai/Kimi-K2-Instruct	moonshotai/kimi-k2-instruct (new lab entry)
Qwen/Qwen3.8-Flash-Next	alibaba/qwen3.8-flash-next
Qwen/Qwen3.8-27B	alibaba/qwen3.8-27b
abacusai/Smaug-Flash	abacusai/smaug-flash (new lab entry; Abacus.AI finetune of DeepSeek V4 Flash, https://abacus.ai/help/platform-updates/2026-08-01-platform-update)

Scope notes

• STT/transcription SKUs (gpt-realtime-whisper, gpt-transcribe, gpt-4o-transcribe, gpt-4o-mini-transcribe) are again excluded, consistent with the review decision in feat(abacus): add missing text-generation models from RouteLLM catalog #4705.
• Image/audio/video-generation SKUs (image gen, TTS, video models) are out of scope, matching feat(abacus): add missing text-generation models from RouteLLM catalog #4705.
• Provider files are strictly override-only (no restated lab fields); partial limit differences are single-key overrides (deep-merge).
• Where RouteLLM does not report thinking for an ID whose lab model is a reasoner, reasoning = true + reasoning_options = [] is set with a header comment (same pattern as zai-org/GLM-5.2).

Validation

• bun validate passes
• git diff --check passes

Source

• https://routellm.abacus.ai/v1/models (accessed 2026-09-23)

[github-actions]

Action items

• [high] [violation] providers/abacus/models/Qwen/Qwen3.8-Flash-Next.toml:11 - Check: Override-only provider files must not restate identical base values. Why: reasoning = true duplicates models/alibaba/qwen3.8-flash-next.toml:19 (reasoning = true) with no delta; RouteLLM comment confirms no thinking difference. Action: Delete the reasoning = true line and inherit from base_model.
• [high] [violation] providers/abacus/models/abacusai/Smaug-Flash.toml:11 - Check: Override-only provider files must not restate identical base values. Why: reasoning = true duplicates new models/abacusai/smaug-flash.toml:75 (reasoning = true). Action: Delete the reasoning = true line.
• [high] [violation] providers/abacus/models/claude-opus-5-5-thinking.toml:11 - Check: Override-only provider files must not restate identical base values. Why: reasoning = true duplicates models/anthropic/claude-opus-5-5.toml:7 (reasoning = true). Action: Delete the reasoning = true line.
• [high] [violation] providers/abacus/models/grok-4.7.toml:11 - Check: Override-only provider files must not restate identical base values. Why: reasoning = true duplicates models/xai/grok-4.7.toml:9 (reasoning = true); the real deltas are already expressed via [limit]/[modalities]/cost. Action: Delete the reasoning = true line.
• [high] [violation] providers/abacus/models/mimo-v2.6-pro.toml:11 - Check: Override-only provider files must not restate identical base values. Why: reasoning = true duplicates models/xiaomi/mimo-v2.6-pro.toml:7 (reasoning = true). Action: Delete the reasoning = true line.
• [high] [violation] providers/abacus/models/zai-org/GLM-5.3.toml:11 - Check: Override-only provider files must not restate identical base values. Why: reasoning = true duplicates models/zhipuai/glm-5.3.toml:7 (reasoning = true). Action: Delete the reasoning = true line.
• [high] [violation] providers/abacus/models/zai-org/GLM-5.3-Flash.toml:11 - Check: Override-only provider files must not restate identical base values. Why: reasoning = true duplicates models/zhipuai/glm-5.3-flash.toml:8 (reasoning = true). Action: Delete the reasoning = true line.
• [high] [violation] providers/abacus/models/moonshotai/Kimi-K2-Instruct.toml:16 - Check: Override-only provider files must keep only real deltas. Why: [limit] output = 65_536 is identical to new models/moonshotai/kimi-k2-instruct.toml:16; context 262_144 is already inherited so the whole table is redundant. Action: Remove the [limit] table entirely.
• [medium] [possible mistake] providers/abacus/models/gpt-6-astra.toml:10 - Check: Inherited limit.input must not go stale when overriding limit.context. Why: Base has context = 1_050_000 + input = 922_000; file overrides context = 1_000_000 for RouteLLM context_length: 1000000 with no input source, leaving inherited 922_000 via deep-merge. Established peers (providers/abacus/models/gpt-5.4.toml:11, gpt-5.6-sol.toml:11) use base_model_omit = ["limit.input"] for this case. Action: Add base_model_omit = ["limit.input"] or verify/provide a RouteLLM input source.
• [medium] [possible mistake] providers/abacus/models/gpt-6-sol.toml:10 - Check: Inherited limit.input must not go stale when overriding limit.context. Why: Same as gpt-6-astra: base 1_050_000/922_000 vs RouteLLM 1000000, no omit, stale 922_000 retained. Action: Add base_model_omit = ["limit.input"] or verify/provide a RouteLLM input source.
• [medium] [possible mistake] providers/abacus/models/gpt-6-luna.toml:10 - Check: Inherited limit.input must not go stale when overriding limit.context. Why: Same as gpt-6-astra: base 1_050_000/922_000 vs RouteLLM 1000000, no omit, stale 922_000 retained. Action: Add base_model_omit = ["limit.input"] or verify/provide a RouteLLM input source.

[github-actions]

No actionable findings.

[sorenjuul]

@rekram1-node new models added since #4705 so they are available for users of abacus

[github-actions]

No actionable findings.

[github-actions]

No actionable findings.

[github-actions]

No actionable findings.

[github-actions]

No actionable findings.

[github-actions]

No actionable findings.

[github-actions]

No actionable findings.

[github-actions]

No actionable findings.

[sorenjuul]

@rekram1-node this PR will enable us to use a lot of new models - hope you will take a look

[rekram1-node]

rlly none of the models allow you to specify reaosning efforts or anything>???

[github-actions]

Closing this pull request as stale because it has not been updated since feedback from @rekram1-node was left 7 days ago. Feel free to reopen it or submit a new pull request if the work is resumed.