Leaderboard / Liquid AI
LFM3-Omni 2B
Liquid AI · LFM3 · released 2026-10-21 · data checked 2026-11-01
LFM3-Omni 2B is a 2.2B-parameter dense model from Liquid AI (United States), released 2026-10-21 under the LFM Open License license. It doesn't have enough reported benchmarks for an index score yet. It needs about 2 GB of memory at 4-bit with an 8K-token context.
Takes text, image, audio and video; answers in text or speech.
Benchmarks
Each bar shows this model's score. The tick marks the best tiny model on that benchmark. Sources: 1 reported by Liquid AI. How we label sources.
Specifications
- Lab
- Liquid AI (US · United States)
- Architecture
- Dense transformer
- Total parameters
- 2.2B
- Active parameters
- All (dense)
- Size class
- Tiny
- Context window
- 128K tokens
- Input modalities
- text, image, audio, video
- Reasoning
- Instruct (no thinking)
- License
- LFM Open License
- Openness
- Open weights
- Memory to run
- ~2 GB at 4-bit, 8K context (more sizes)
- Hugging Face
- 11.7K downloads (30 days) · 324 likes
Compare with
- LFM3-Omni 2B vs Qwen3.5 2B (2.3B, index 33.0)
- LFM3-Omni 2B vs MiniCPM5 2B (2.5B, index 57.5)
- LFM3-Omni 2B vs LFM2.5 2.6B (2.7B, index 43.2)
Sources
Something wrong here? Report it.
Run it locally
Uses LiquidAI/LFM3-Omni-2B-GGUF (Q4_K_M). Swap the tag for another size: BF16, F16, Q4_0, Q4_K_M, Q8_0.
ollama run hf.co/LiquidAI/LFM3-Omni-2B-GGUF:Q4_K_Mllama-server -hf LiquidAI/LFM3-Omni-2B-GGUF:Q4_K_Mpip install -U mlx-lm
mlx_lm.chat --model LiquidAI/LFM3-Omni-2B-MLX-4bitvllm serve LiquidAI/LFM3-Omni-2BMemory needed: weights, context cache and runtime buffers.
| Quantization | 4K context | 32K context | 128K context |
|---|---|---|---|
| 16-bit F16 | 4.9 GB | 5.3 GB | 6.5 GB |
| 8-bit Q8_0 | 2.9 GB | 3.3 GB | 4.4 GB |
| 4-bit Q4_K_M | 1.9 GB | 2.3 GB | 3.4 GB |