2026.3.0.0

What's Changed

• Bump product version 2026.3.0 by @akladiev in #3858
• [wwb] Update video inputs to avoid big video by @sbalandi in #3777
• Bump the npm-dependencies group across 1 directory with 2 updates by @dependabot[bot] in #3849
• Bump typescript-eslint from 8.59.2 to 8.59.3 in /site in the npm-dependencies group across 1 directory by @dependabot[bot] in #3850
• skip test_qwen3_embedding_npu tests by @as-suvorov in #3859
• [samples] Move encrypted_model_vlm from minicpmv to qwen3 by @sbalandi in #3842
• Bump the pip-dependencies group across 5 directories with 6 updates by @dependabot[bot] in #3824
• [llm_becn/wwb] Update torch/torchvision version range by @sbalandi in #3863
• Don't require ABI stability by @Wovchena in #3861
• Fix Gemma3 chat template formatting by @yatarkan in #3867
• [C API] ChatHistory support in VLM pipeline by @zhaohb in #3631
• Bump openvinotoolkit/openvino from 7f61dc3cf40c03638386397c6f8f98bf23f25a00 to 48b684f1ba3a01e573b3de6444e9e37756e3cdc9 in the github-actions group across 1 directory by @dependabot[bot] in #3851
• Bump the pip-dependencies group across 5 directories with 7 updates by @dependabot[bot] in #3869
• Bump the pip-dependencies group across 3 directories with 2 updates by @dependabot[bot] in #3878
• Bump brace-expansion from 5.0.5 to 5.0.6 in /.github/actions/install_wheel by @dependabot[bot] in #3880
• Bump openvinotoolkit/openvino from 48b684f1ba3a01e573b3de6444e9e37756e3cdc9 to 5db1bfbd991e4f604111d9c69b0007f5a2f8f87c in the github-actions group by @dependabot[bot] in #3877
• Bump webpack-dev-server from 5.2.3 to 5.2.4 in /site by @dependabot[bot] in #3882
• [JS] Add VLMPipeline samples by @Retribution98 in #3287
• Bump typescript-eslint from 8.59.3 to 8.59.4 in /site in the npm-dependencies group across 1 directory by @dependabot[bot] in #3884
• [JS] Fix CI fails by @Retribution98 in #3879
• Bump the pip-dependencies group across 2 directories with 1 update by @dependabot[bot] in #3886
• Bump typescript-eslint from 8.59.3 to 8.59.4 in /src/js in the npm-dependencies group across 1 directory by @dependabot[bot] in #3883
• Bump the github-actions group with 2 updates by @dependabot[bot] in #3885
• Deprecated start_chat() / finish_chat() API in VLM and CB pipelines by @yatarkan in #3889
• [Docs] Fix build by @yatarkan in #3891
• Kokoro Text-to-Speech Support by @RyanMetcalfeInt8 in #3483
• [WWB] Create gt and output folders if they don't exist by @CuriousPanCake in #3890
• Decrease GGUF timeouts by @Wovchena in #3894
• Bump optimum-intel to May 20 (140e49a6) for all tests by @peterchen-intel in #3852
• Remove deprecation prints from C++ by @yatarkan in #3899
• Add denoising_process C++ and Python samples by @byekrang in #3881
• Skip Mac GGUF tests by @Wovchena in #3902
• [SD] Dynamic Tree Search Sampler for draft generation and main validation by @GuoliangShiIntel in #3451
• Bump the github-actions group across 1 directory with 3 updates by @dependabot[bot] in #3892
• Bump langchain-community from 0.4.1 to 0.4.2 in /tests/python_tests in the pip-dependencies group across 1 directory by @dependabot[bot] in #3908
• Bump typescript-eslint from 8.59.4 to 8.60.0 in /site in the npm-dependencies group across 1 directory by @dependabot[bot] in #3912
• [WWB] Skip prompt with incoherent answer for Qwen3.5-0.8B by @yatarkan in #3898
• Bump typescript-eslint from 8.59.4 to 8.60.0 in /src/js in the npm-dependencies group across 1 directory by @dependabot[bot] in #3910
• Bump the github-actions group across 1 directory with 2 updates by @dependabot[bot] in #3913
• Bump eslint-plugin-prettier from 5.5.5 to 5.5.6 in /src/js in the npm-dependencies group across 1 directory by @dependabot[bot] in #3918
• [llm_bench] Add per-stage performance metrics for WhisperPipeline and sampling duration for LLMPipeline by @goyaladitya05 in #3669
• Bump openvinotoolkit/openvino from 51b2fef55c075a40ad3a5e84dea1ffbf68f88b5f to ac001bacd3a9f207db30326a605154e14783b40a in the github-actions group by @dependabot[bot] in #3919
• Bump typescript-eslint from 8.60.0 to 8.60.1 in /src/js in the npm-dependencies group across 1 directory by @dependabot[bot] in #3924
• Added dense LFM2 class to text_gen type by @AsyaPronina in #3929
• Bump the npm-dependencies group across 1 directory with 4 updates by @dependabot[bot] in #3926
• Fix embedding start time for VLM pipeline on NPU by @JohnLeFeng in #3771
• Replace skipif() back with xfail() by @Wovchena in #3905
• [wwb] make long prompt by default by @sbalandi in #3903
• [CI] Skip Kokoro and Gemma3 (PA) failing tests by @yatarkan in #3945
• Bump the github-actions group across 1 directory with 2 updates by @dependabot[bot] in #3925
• Bump the npm-dependencies group across 1 directory with 2 updates by @dependabot[bot] in #3952
• Bump the pip-dependencies group across 3 directories with 6 updates by @dependabot[bot] in #3949
• Bump typescript-eslint from 8.60.1 to 8.61.0 in /site in the npm-dependencies group across 1 directory by @dependabot[bot] in #3953
• [wwb/llm_bench] Add skills for fail analysis by @sbalandi in #3896
• [CI] [GHA] Use ninja as a generator for wheel build on mac by @akashchi in #3928
• [wwb, llm_bench] Move to dtype auto for pt models by @sbalandi in #3793
• ignore cache_interval_multiplier for non LM models by @dtrawins in #3943
• MODEL_PROPERTIES support for all VLM model roles by @dkalinowski in #3781
• wwb: Add support for Kokoro by @RyanMetcalfeInt8 in #3897
• add truncation genai::Tokenizer::encode by @pavel-esir in #3876
• Bump shell-quote from 1.8.3 to 1.8.4 in /site by @dependabot[bot] in #3959
• Bump @types/node from 24.13.1 to 24.13.2 in /src/js in the npm-dependencies group across 1 directory by @dependabot[bot] in #3969
• Bump the github-actions group across 1 directory with 2 updates by @dependabot[bot] in #3970
• Add LFM2(.5) and Qwen3Next to the Docs by @apaniukov in #3972
• Kokoro tts test fix by @RyanMetcalfeInt8 in #3961
• Bump the pip-dependencies group across 2 directories with 3 updates by @dependabot[bot] in #3971
• [CMAKE] Add pdb install by @mryzhov in #3973
• Fix StaticWhisperPipeline to work with transformers v5 by @eshiryae in #3923
• Bump the github-actions group with 2 updates by @dependabot[bot] in #3976
• [wwb] add process inputs for gemma-4 by @sbalandi in #3960
• Bump the pip-dependencies group across 3 directories with 5 updates by @dependabot[bot] in #3983
• ContinuousBatching Pipeline Leftovers by @apaniukov in #3854
• Bump launch-editor from 2.13.2 to 2.14.1 in /site by @dependabot[bot] in #3985
• [JS] Implement Image2ImagePipeline and InpaintingPipeline by @Retribution98 in #3916
• Bump js-yaml from 4.1.1 to 4.2.0 in /src/js by @dependabot[bot] in #3986
• Bump ws from 7.5.10 to 7.5.11 in /site by @dependabot[bot] in #3984
• Bump typescript-eslint from 8.61.0 to 8.61.1 in /src/js in the npm-dependencies group across 1 directory by @dependabot[bot] in #3987
• Bump the github-actions group across 1 directory with 2 updates by @dependabot[bot] in #3988
• Bump the npm-dependencies group across 1 directory with 2 updates by @dependabot[bot] in #3995
• Bump webpack-dev-server from 5.2.4 to 5.2.5 in /site by @dependabot[bot] in #3996
• Bump the github-actions group with 2 updates by @dependabot[bot] in #3999
• Bump undici from 6.25.0 to 6.27.0 in /site by @dependabot[bot] in #4005
• Support gcc15 by @as-suvorov in #4006
• Bump undici and cheerio in /site by @dependabot[bot] in #4007
• Bump typescript-eslint from 8.61.1 to 8.62.0 in /src/js in the npm-dependencies group across 1 directory by @dependabot[bot] in #4010
• Add ASRPipeline interface by @as-suvorov in #3893
• Bump the pip-dependencies group across 5 directories with 7 updates by @dependabot[bot] in #4004
• Enable eagle3 top1 in vlm by @sunxiaoxia2022 in #3330
• [SD] eagle3 cb backend pipeline enable on top of dynamic tree drafting by @songbell in #3755
• CVS-175243 use primary hf token openvino by @jkoniusz in #3818
• Fix prefix cache issue for eagle3 by @xufang-lisa in #3887
• [wwb] Force dtype float32 for models loaded with diffusers by @sbalandi in #4016
• [skills]: Add vlm model by @as-suvorov in #3759
• Update third-party-programs.txt by @ValentinaKats in #4035
• Update pandas requirement from <=3.0.3,>=2.0.3 to >=2.0.3,<=3.0.4 in /tools/who_what_benchmark in the pip-dependencies group across 1 directory by @dependabot[bot] in #4052
• [JS] Update Node.js README by @Retribution98 in #3909
• Fix CVS-189485 by @sunxiaoxia2022 in #4032
• Kczarnec malloc trim in memory monitor cooldown by @krzyczar in #3803
• [VLM] Support PA for Gemma4 by @yatarkan in #3994
• Fix heap buffer overflow in GGUF Q4_K/Q6_K dequantizer by @TianmengChen in #4002
• Bump typescript-eslint from 8.62.0 to 8.62.1 in /src/js in the npm-dependencies group across 1 directory by @dependabot[bot] in #4057
• [JS] Add decode for image generation pipelines by @Retribution98 in #4013
• Qwen3-ASR by @as-suvorov in #3991
• [wwb] Add trimming kv_cache for vlm/llm chat pipelines by @sbalandi in #3998
• [CB] Fix high VRAM and empty-output for hybrid SSM models with prefix caching (#3968, #4047) by @exzile in #4050
• align prompt chunks for eagle3 by @xufang-lisa in #3812
• [llm_bench/wwb] Add more information about ov configuration to tools documentation by @sbalandi in #4054
• VIT optimization on VideoChat-Flash by @WeldonWangwang in #3756
• add dtor for ThreadedCallbackWrapper by @pavel-esir in #3932
• Gemma4 MTP Stateful Pipeline Support by @apaniukov in #3958
• Reject empty vocab in structured output constructor by @mzegla in #4062
• [wwb] Skip visual-text due to default dataset is broken by @sbalandi in #4072
• [wwb] Skip test_pruner_basic due to default dataset for visual-text is broken by @sbalandi in #4073
• Add gemma-4-12b-it by @as-suvorov in #4001
• [NPU] Eagle 3 Topk Pipeline by @GuoliangShiIntel in #3396
• [SD]support tree sd tests in llm_bench and wwb by @songbell in #4053
• [CB] Reserve dynamic cache capacity for later/larger prompts in one reallocation (#3968) by @exzile in #4051
• Bump the pip-dependencies group across 5 directories with 3 updates by @dependabot[bot] in #4067
• Add Gemma3n model by @as-suvorov in #4075
• Bump the github-actions group across 1 directory with 7 updates by @dependabot[bot] in #4033
• [wwb] Move dataset for VLM instruction cases to lmms-lab/VQAv2 by @sbalandi in #4071
• WWB: add llama.cpp chat and n_ctx CLI support with validation and regression tests by @rasapala in #3974
• Update actions and js dependencies by @as-suvorov in #4086
• Bump typescript-eslint from 8.62.1 to 8.63.0 in /src/js in the npm-dependencies group across 1 directory by @dependabot[bot] in #4098
• Bump the pip-dependencies group across 5 directories with 4 updates by @dependabot[bot] in #4099
• [Fix Coverity] Dereference of potentially null field by @GuoliangShiIntel in #4087
• Bump the github-actions group across 1 directory with 6 updates by @dependabot[bot] in #4088
• Gemma4 unified apply wwb add_bos_token fix by @as-suvorov in #4100
• llm_bench: Add initial support for Kokoro TTS model by @RyanMetcalfeInt8 in #4020
• [JS] Bump NPM package to 2026.3 by @Retribution98 in #4103
• [wwb] Update extras llm-test-openvino to align with new versions by @sbalandi in #4104
• [ImageGeneration] Flux2Klein support by @likholat in #3930
• fix final parser m_expect_open_tag = false by @pavel-esir in #3956
• Fix xgrammar structured-output crash at EOS under speculative decoding by @blairducrayoppat in #4082
• Bump tqdm from 4.68.3 to 4.68.4 in /tools/who_what_benchmark in the pip-dependencies group across 1 directory by @dependabot[bot] in #4109
• [JS] Update Node Addon API version by @Retribution98 in #4106
• Enable CDPruner vision token pruning for Qwen3-VL by @yangwang201911 in #3499
• [Tests] Unskip GenAI vs Optimum tests by @yatarkan in #3981
• Smart-ci deps workaround by @as-suvorov in #4118
• [llm_bench] Resolving complicated use cases moved from tool to --task by @sbalandi in #4063
• [Release] Add EmbeddingPipeline class for Qwen3-VL-Embedding (#4093) by @pavel-esir in #4145
• Adjust KV cache capacity calculation with NPUW plugin by @AlexanderKalistratov in #4147
• [Release] Cache Minja Template by @apaniukov in #4164
• [VLM] QWEN-3-Omni support to RELEASE 2026.03 by @sgonorov in #4166
• [llm_bench] Restore Python < 3.11 compatibility for av by @wkobielx in #4184
• [Release] Fix double video frames sampling for Qwen3-VL and Qwen3-Omni by @yatarkan in #4207

New Contributors

• @byekrang made their first contribution in #3881
• @ValentinaKats made their first contribution in #4035
• @exzile made their first contribution in #4050
• @blairducrayoppat made their first contribution in #4082

Full Changelog: 2026.2.1.0...2026.3.0.0