Back to changelog
Gateway
September 1, 2026
New models, text-to-video, and hosted tool_search support
The model catalog gains Claude Fable 5.1, Gemini 3.8 Flash, and text-to-video through fal.ai
New
Merge now supports Wafer as a new model vendor, plus extra provider routes for four existing models.
- Claude Fable 5.1 (Anthropic) and Gemini 3.8 Flash (Google, zero data retention) are now supported
- Text-to-video generation runs through fal.ai, including models such as Seedance 2.5
- Wafer is available as a model vendor, with zero data retention
- DeepSeek V4 Pro 0813, MiniMax M3, and Qwen3.8 Max now route through Fireworks AI, and GLM-5.3 Flash through Baseten, Particle AI, and Wafer
Read more in the Gateway docs.
Responses endpoint supports OpenAI's hosted tool_search
Improved
Codex can now complete its deferred tool-loading round trip through Gateway. Read more in the Gateway docs.
Improvements
- Logs - Adds a rollup summary strip with tokens, cache activity, and estimated cost
- Logs and Security - Timestamps now display in the viewer's local timezone
- Model Explorer - Surfaces promotional pricing with the standard price struck through and the revert date
- Model Explorer, /v1/models - Shows long-context pricing on list, detail, and pricing-block views
- Usage and Spend - Adds a By Routing Policy tab with spend, token rollups, and routed requests
- Web search - Interfaze is now supported as a search engine, backed by the OpenWebSearch API