What's Changed
- feat(server): live-input STT WebSocket on the audio worker by @TheAngryPit in #2167
- Add SoMark OXR-1.0 document parsing model by @bxespinozar in #2560
- Reduce CI turnaround with sharding and cheaper parity tests by @Blaizzy in #2567
- fix: keep converted metadata writable by @ykhrustalev in #2298
- fix: record converted dtype in exported configs by @Blaizzy in #2577
- Fix text-only input truncation by @Mengmj in #2552
- feat: native Nemotron 3.5 Parse MLX port by @geoHeil in #2562
- Fix ModelOpt NVFP4 AWQ checkpoint loading and input scaling by @Blaizzy in #2582
- Expose generation timings in Responses API results by @Blaizzy in #2583
- perf: make d1 and d1-omni faster and use less memory by @ykhrustalev in #2500
- Add MiniMax M3.1 Flash support by @Blaizzy in #2580
- (feat) Add mtp to trainers by @Goekdeniz-Guelmez in #2558
- Add streamed diarized transcriptions and speaker analysis endpoints by @Blaizzy in #2600
- Bump version to 0.7.11 by @lucasnewman in #2599
- docs: list Granite Vision 4.1 and Mistral-Small-4 in model READMEs by @DineshKadiam in #2598
- Recover compaction failures and reuse chat prefixes by @Blaizzy in #2602
- Add Unlimited-OCR-2 support (page2d RoPE + DASH sparse attention) by @shuuul in #2601
- Add native LFM3 and LFM3-VL support by @Blaizzy in #2607
- Align LFM3 video and reasoning with release checkpoints by @Blaizzy in #2608
- Add LFM3-Audio support by @Blaizzy in #2609
- Add LFM3-Omni support by @Blaizzy in #2610
- Add LFM3 d1 decision models by @Blaizzy in #2611
- Add MiniCPM-OCR 1.0 and Jev support by @Blaizzy in #2606
- Fix Unlimited-OCR FP32 fusion and chunked prefill reference retention by @bearcatty in #2603
- Forward native STT prompts and task selection in the audio API by @Blaizzy in #2618
- Exclude transient request context from compaction history by @Blaizzy in #2612
- Exclude tool-call examples in reasoning from executable calls by @Blaizzy in #2613
- Recover long tool conversations and reduce cache overhead by @Blaizzy in #2617
New Contributors
- @TheAngryPit made their first contribution in #2167
- @bxespinozar made their first contribution in #2560
- @Mengmj made their first contribution in #2552
- @geoHeil made their first contribution in #2562
- @DineshKadiam made their first contribution in #2598
- @bearcatty made their first contribution in #2603
Full Changelog: v0.7.10...v0.7.11