Model catalog update — 2026-09-29

Weekly automated catalog run. Every row below was read from the provider's own catalogue or pricing page, answered a live call, and was reviewed and merged by a human before release.

Stable model names: flash and think

model: "flash" (cheapest current model) and model: "think" (cheapest current high-quality model measured to reason) resolve per region and PHI class and follow new releases, price changes and shutdowns automatically. GET /v1/models shows what each resolves to today.

New models

• gemini-3-1-flash-lite — Gemini 3.1 Flash Lite (Google): $0.29 / $1.74 per 1M tokens, chat + vision, tools, 1,048,576 context / 65,536 output.
• gemini-3-5-flash-lite — Gemini 3.5 Flash Lite (Google): $0.35 / $2.91 per 1M tokens, chat + vision, tools, 1,048,576 context / 65,536 output.
• scaleway-qwen3-6-35b now also serves code (Scaleway's designated successor to qwen3-coder-30b).

Upstream shutdowns

Models leave auto routing and the default model list 31 days before their vendor's announced shutdown; a pinned model keeps working until the date itself.

• scaleway-qwen3-coder-30b — Scaleway end-of-life 2026-10-01.
• gemini-2-5-flash, gemini-2-5-pro — Google retirement 2026-10-20.
• gpt-5, gpt-5-mini — OpenAI shutdown 2026-12-11.

Scheduled price changes

Google's introductory prices end on 2027-01-01. The new rates apply automatically on that date:

• gemini-3-7-flash, gemini-3-8-flash: $0.87 / $4.36 → $1.74 / $8.72 per 1M tokens.
• gemini-3-8-flash-tts: $0.58 / $10.46 → $1.16 / $20.93; gemini-3-8-flash-lite-tts: $0.58 / $6.98 → $1.16 / $13.95.

Default changes

Unpinned (model: "auto") traffic resolves differently for:

• chat + vision / UK, ASIA, ANZ, LATAM, MENA, INDIA, WORLD / general: gemini-2-5-flash → gemini-3-1-flash-lite
• chat + vision / WORLD / general: → mistral-small (WORLD now picks the cheapest model across all of its providers; think → gpt-5)
• chat / EU / general + phi: scaleway-qwen3-coder-30b → scaleway-gemma-4-26b
• code / EU / general + phi: scaleway-qwen3-coder-30b → scaleway-qwen3-6-35b

Added

• Claude Sonnet 5.5 (claude-sonnet-5-5) — $2.33 / $11.63 per 1M tokens, US/EU/ASIA/WORLD, chat + vision, tools; context 1,000,000, max output 128,000.

Archived (older generations)

Still fully routable when pinned; hidden from /v1/models and /v1/pricing unless ?include_archived=true.

• claude-sonnet-4-6

---

Later run

Weekly automated catalog run. Every row below was read from the provider's own catalogue or pricing page, answered a live call, and passed every routing check before it went live.

Added

• gpt-5-nano (gpt-5-nano) — $0.06 / $0.47 per 1M tokens, US/EU/WORLD, chat + vision, tools; context 400,000, max output 128,000.
• gpt-5.4 (gpt-5-4) — $2.91 / $17.44 per 1M tokens, US/EU/WORLD, chat + vision, tools; context 1,050,000, max output 128,000.
• gpt-5.4-mini (gpt-5-4-mini) — $0.87 / $5.23 per 1M tokens, US/EU/WORLD, chat + vision, tools; context 400,000, max output 128,000.
• gpt-5.4-nano (gpt-5-4-nano) — $0.23 / $1.45 per 1M tokens, US/EU/WORLD, chat + vision, tools; context 400,000, max output 128,000.
• gpt-5.5 (gpt-5-5) — $5.81 / $34.88 per 1M tokens, US/EU/WORLD, chat + vision, tools; context 1,050,000, max output 128,000.
• gpt-6-luna (gpt-6-luna) — $0.12 / $0.58 per 1M tokens, US/EU/WORLD, chat + vision; context 1,050,000, max output 128,000.
• gpt-6-sol (gpt-6-sol) — $2.33 / $11.63 per 1M tokens, US/EU/WORLD, chat + vision; context 1,050,000, max output 128,000.
• gpt-6.1-sol (gpt-6-1-sol) — $2.33 / $11.63 per 1M tokens, US/EU/WORLD, chat + vision; context 1,050,000, max output 128,000.

Default changes

Unpinned (model: "auto") traffic resolves differently for:

• chat / WORLD / general: mistral-small → gpt-5-nano
• vision / WORLD / general: mistral-small → gpt-5-nano
• think / WORLD / general: gpt-5 → gpt-6-sol

Archived (older generations)

Still fully routable when pinned; hidden from /v1/models and /v1/pricing unless ?include_archived=true.

• gpt-5
• gpt-5-6-sol

Upstream shutdowns

Models leave auto routing and the default model list 31 days before the vendor's announced shutdown; a pinned model keeps working until the date itself.

• gpt-5-nano (openai) — 2026-12-11, vendor replacement gpt-5.6-luna.

---

Later run

Weekly automated catalog run. Every row below was read from the provider's own catalogue or pricing page, answered a live call, and passed every routing check before it went live.

Added

• mistral-large-latest (mistral-large-latest) — $0.58 / $1.74 per 1M tokens, EU/ASIA/WORLD, chat + vision, tools; context 262,144, max output bounded by context.
• kimi-k2.6 (Moonshot native) (kimi-k2-6) — $1.10 / $4.65 per 1M tokens, US, chat + code + vision, tools; context 262,144, max output bounded by context.
• kimi-k2.7-code (Moonshot native) (kimi-k2-7-code) — $1.10 / $4.65 per 1M tokens, US, chat + code + vision, tools; context 262,144, max output bounded by context.