v2.1.0

New Model Support

• Gemma 3n (#1650 by @anatyrova)
• Qwen3-Omni-MOE (#1700 by @sgonorov)
• SmolLM3 (#1715 by @AshutoshSinghIntel)
• Qwen3-VL-Embedding (#1723 by @mlukasze)
• Gemma 4 Unified (#1770 by @rkazants)
• FLUX.2 (#1809 by @rkazants)

Improvements & fixes

• Transformers v5.1-5.5 support (#1684 by @echarlaix)
• Fix Kokoro voice conversion from local model directory (#1803 by @mmikolajcz)
• Extended OpenVINO transformation verification tests for MoE architectures (#1776 by @anatyrova)
• Unpatch 16-bit models previously patched with patch_16bit_model after conversion (#1811 by @anatyrova)
• Set dynamic sequence length for FLUX.2 (#1846 by @rkazants)
• Deprecated the broken contextual calibration dataset, replaced with textvqa (#1849 by @echarlaix)
• Fixed passing of token_type_ids for Gemma3 (#1881 by @popovaan)
• Fixed export for encoder-decoder models with unnamed tensors (#1857 by @echarlaix)
• Fixed long audio inference for Qwen3-ASR (#1782 by @popovaan)
• Fixed ssm states handling for Mamba and Falcon-Mamba (#1779 by @echarlaix)
• Pinned huggingface_hub<1.22 to fix offline loading (#1855 by @echarlaix)
• Use base transformers version for version comparisons (#1798 by @AlexanderDokuchaev)

Other Changes

• CI: local model caching and rate-limit fixes for fork PRs (#1817, #1826, #1835, #1837, #1840, #1842 by @echarlaix)
• Faster doc builds (#1851 by @mishig25)

New Contributors

• @anatyrova made their first contribution in #1650
• @AshutoshSinghIntel made their first contribution in #1715
• @mlukasze made their first contribution in #1723

Full Changelog

v2.0.0...v2.1.0

Compatible transformers version

Compatible with transformers>=4.51,<5.6

NPU-specific requirements:

• Whisper - need transformers v5.0
• LFM2 - need transformers v5.0

Maximum supported transformers version by model architectures can be found in the table below (models not listed support up to transformers v5.6)

Model architecture	Max transformers version
AquilaM	4.57.6
Arctic	4.53.3
Baichuan	4.57.6
Bitnet	4.57.6
ChatGLM2	4.55.4
DBRX	4.57.6
Data2VecText	4.57.6
Deci	4.57.6
Deepseek	4.53.3
Exaone	4.57.6
Exaone4	4.57.6
Flaubert	4.57.6
Gemma	5.0
Gemma2	5.0
Gemma3Text	5.0
Gemma3nText	5.0
Gemma4Text	5.0
Gemma4Unified	5.10.4
Gemma4UnifiedText	5.10.4
GLM	5.0
GotOCR2	4.57.6
GraniteMoeHybrid	5.3.0
Idefics3	4.57.6
InternLM	4.57.6
InternLM2	4.57.6
InternVLChat	4.57.6
Jais	4.57.6
LFM2	5.4.0
LFM2Moe	5.4.0
Llama4	4.57.6
Llama4Text	4.57.6
LlavaNextVideo	4.57.6
LlavaQwen2	4.53.3
Mamba	5.3.0
Marian	4.57.6
MiniCPM	4.53.3
MiniCPM3	4.53.3
MiniCPMO	4.51.3
MiniCPMV	4.57.6
MT5	4.57.6
Nystromformer	4.50.3
Orion	4.57.6
Phi3Vision	4.53.3
Phi4MM	4.53.3
Qwen	4.55.4
Qwen2VL	5.0
Qwen2_5_VL	5.0
Qwen3ASR	4.57.6
Qwen3Next	4.57.6
Qwen3VL	5.0
Qwen3_5	5.2.0
Qwen3_5Moe	5.2.0
Qwen3_5MoeText	5.2.0
Qwen3_5Text	5.2.0
SmolVLM	4.57.6
VideoChatFlashQwen	4.57.6
XLM	4.57.6
XverseM	4.57.6
Zamba2	4.57.6

Recommended versions

• OpenVINO: v2026.3
• OpenVINO GenAI: v2026.3
• NNCF: v3.3