Small language models, compared
Benchmarks, size, price and capabilities for 139 open-weight models under 100B parameters from 49 labs. Use it to pick a model you can run on one GPU, a laptop or a phone, or one that's cheap to serve.
Best Tiny (< 5B)
70.4
Nanbeige4.2 3B · 4.2B
Best Small (5–15B)
63.9
Qwen3.5 9B · 9B
Best Medium (15–40B)
77.8
Nanbeige4.3 33B · 33.1B
Best Large (40–100B)
73.3
Nemotron 3.5 Bolt 74B-A9B · 74B
Best for: Coding · Math · 8 GB GPU / laptop · 16 GB GPU · 24 GB GPU · Phones and edge · Cheap API · Vision (image input) · Apache / MIT license · Made in Europe · Made in the US · Fully open
Compare on any two axes
Pick what goes on each axis: the index, any benchmark, size, memory, price, context, adoption or release date. The dashed line joins models nothing else beats on both axes. Click a point to open the model.
Frontier: nothing beats these on both axes (labelled)87 plotted · 52 have no data for these axes
Leaderboard
139 models · Prices are USD per 1M tokens (input / output), cheapest OpenRouter routeCompare 0/4
| Select | # | Input | License | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Nanbeige4.3 33BNanbeige (Kanzhun) · CN | 77.82/6 | — | 89.5 | 95.2 | — | — | 33.1B | dense | 524K | — | 2026-10-24 | T | Apache-2.0 | |
| 2 | Qwen3.8 27BAlibaba (Qwen) · CN | 77.03/6 | — | 89.2 | — | 90.3 | — | 27B | dense | 262K | $0.009 / $0.49 | 2026-08-14 | T I V | Apache-2.0 | |
| 3 | Ternary Bonsai 2 27BPrismML · US | 74.62/6 | — | — | 95.8 | 90.1 | — | 27.4B | dense | 262K | — | 2026-09-17 | T I | Apache-2.0 | |
| 4 | Qwen3.6 27BAlibaba (Qwen) · CN | 74.06/6 | 86.2 | 87.8 | 94.1 | 83.9 | 77.2 | 27B | dense | 262K | $0.20 / $3.00 | 2026-04-22 | T I V | Apache-2.0 | |
| 5 | Nemotron 3.5 Bolt 74B-A9BNVIDIA · US | 73.32/6 | — | 88.8 | — | — | — | 74B | 9B | 262K | — | 2026-10-12 | T I | OpenMDW-1.1 | |
| 6 | Muse Glimmer 30BMeta · US | 72.64/6 | — | 83.5✓ | 94.7 | — | 76.0 | 29.6B | dense | 131K | $0.30 / $1.20 | 2026-08-09 | T I | Apache-2.0 | |
| 7 | Qwen3.6 35B-A3BAlibaba (Qwen) · CN | 72.26/6 | 85.2 | 86.0 | 92.7 | 80.4 | 73.4 | 35B | 3B | 262K | $0.010 / $0.092 | 2026-04-16 | T I V | Apache-2.0 | |
| 8 | Qwen3.5 27BAlibaba (Qwen) · CN | 70.66/6 | 86.1 | 85.5 | 92.6 | 80.7 | 75.0 | 27B | dense | 262K | $0.25 / $2.00 | 2026-02-24 | T I V | Apache-2.0 | |
| 9 | Nanbeige4.2 3BNanbeige (Kanzhun) · CN | 70.44/6 | — | 87.4 | — | 72.5 | 63.6 | 4.2B | dense | 262K | — | 2026-07-21 | T | Apache-2.0 | |
| 10 | Gemma 4 31BGoogle · US | 70.16/6 | 85.2 | 84.3 | 89.2 | 80.0 | 52.0* | 30.7B | dense | 256K | $0.10 / $0.34 | 2026-04-02 | T I | Apache-2.0 | |
| 11 | Qwen3.5 35B-A3BAlibaba (Qwen) · CN | 69.96/6 | 85.3 | 84.2 | 91.0 | 74.6 | 70.0 | 35B | 3B | 262K | $0.065 / $0.75 | 2026-02-24 | T I V | Apache-2.0 | |
| 12 | Kolibri-1Aleph Alpha · DE | 68.36/6 | 80.0 | 84.3 | 96.0 | 85.9 | 66.4 | 78.1B | 3.5B | 1M | — | 2026-10-03 | T | Apache-2.0 | |
| 13 | Spark-X2.6 4BSparkLLM (XHToken) · CN | 67.54/6 | — | 77.1 | 91.2 | — | 62.8 | 4.1B | dense | 1M | — | 2026-10-19 | T | Apache-2.0 | |
| 14 | Gemma 4 26B-A4BGoogle · US | 67.56/6 | 82.6 | 82.3 | 88.3 | 77.1 | 17.4* | 25.2B | 3.8B | 256K | $0.040 / $0.15 | 2026-04-02 | T I | Apache-2.0 | |
| 15 | Granite 4.2 30BIBM (Granite) · US | 67.04/6 | 77.6 | 66.4 | 89.2 | 75.8 | — | 30B | dense | 131K | — | 2026-08-07 | T | Apache-2.0 | |
| 16 | GLM-4.7-FlashZ.ai (GLM) · CN | 66.15/6 | — | 75.2 | 91.6 | 64.0 | 59.2 | 31B | 3B | 200K | $0.061 / $0.40 | 2026-01-19 | T | MIT | |
| 17 | Spark-X2.5 4BSparkLLM (XHToken) · CN | 65.54/6 | — | 73.7 | 90.7 | — | 41.6 | 4.1B | dense | 1M | — | 2026-08-24 | T | Apache-2.0 | |
| 18 | gpt-oss-20bOpenAI · US | 64.06/6 | 75.0* | 71.5 | 91.7 | 61.0* | 60.7 | 20.9B | 3.6B | 131K | $0.011 / $0.045 | 2025-08-05 | T | Apache-2.0 | |
| 19 | Qwen3.5 9BAlibaba (Qwen) · CN | 63.93/6 | 82.5 | 81.7 | — | 65.6 | — | 9B | dense | 262K | $0.025 / $0.036 | 2026-03-02 | T I V | Apache-2.0 | |
| 20 | Nemotron 3.5 LightningNVIDIA · US · unverified | 63.63/6 | 81.9 | 75.4 | — | — | 51.6 | 31.6B | 3.6B | 1M | $0.025 / $0.12 | 2026-08-11 | T | OpenMDW-1.1 | |
| 21 | Granite 4.2 8BIBM (Granite) · US | 63.54/6 | 74.0 | 64.1 | 86.7 | 73.2 | — | 8B | dense | 131K | $0.060 / $0.15 | 2026-08-07 | T | Apache-2.0 | |
| 22 | Qwen3-Next 80B-A3B ThinkingAlibaba (Qwen) · CN | 63.34/6 | 82.7 | 77.2 | 87.8 | 68.7 | — | 80B | 3B | 262K | — | 2025-09-11 | T | Apache-2.0 | |
| 23 | Edge0 35B-A3B (preview)Edge0 · ? | 62.83/6 | 81.0 | 79.8 | 86.6 | — | — | 34.7B | 3B | 262K | — | 2026-09-08 | T I | Apache-2.0 | |
| 24 | Nemotron 3 Nano 30B-A3BNVIDIA · US | 62.66/6 | 78.3 | 73.0 | 89.1 | 68.3 | 38.8 | 31.6B | 3.5B | 1M | $0.060 / $0.24 | 2025-12-15 | T | NVIDIA Open Model | |
| 25 | EXAONE 4.0 32BLG AI Research (EXAONE) · KR | 62.04/6 | 81.8 | 75.4 | 85.3 | 66.7 | — | 32B | dense | 131K | — | 2025-07-15 | T | EXAONE License (non-commercial) | |
| 26 | Sarvam 30BSarvam AI · IN · unverified | 61.63/6 | 80.0 | 66.5 | 88.3 | — | — | 32.2B | 2.4B | 131K | — | 2026-03-03 | T | Apache-2.0 | |
| 27 | LFM2.5 24B-A2BLiquid AI · US | 61.43/6 | — | 74.4 | — | — | 51.0 | 24B | 2B | 128K | $0.050 / $0.70 | 2026-10-13 | T | LFM Open License | |
| 28 | Apriel 1.6 15B ThinkerServiceNow (Apriel) · US | 60.94/6 | 79.0 | 73.0 | 88.0 | — | — | 14.9B | dense | 262K | — | 2025-12-04 | T I | MIT | |
| 29 | Seed-OSS 36B InstructByteDance Seed · CN | 60.36/6 | 82.7 | 71.4 | 84.7 | 67.4 | 56.0 | 36.2B | dense | 512K | — | 2025-08-20 | T | Apache-2.0 | |
| 30 | LFM3 1.7BLiquid AI · US | 60.04/6 | — | 63.4 | — | 73.5 | 43.4 | 1.7B | dense | 262K | — | 2026-10-27 | T | LFM Open License | |
| 31 | Nemotron Labs Dory 1.0 4BNVIDIA · US | 60.03/6 | — | 69.2 | — | — | 50.3 | 4B | dense | 262K | — | 2026-10-16 | T | OpenMDW-1.1 | |
| 32 | Ministral 3 14BMistral AI · FR | 60.03/6 | — | 71.2 | 85.0 | 64.6 | — | 14B | dense | 256K | $0.20 / $0.20 | 2025-12-02 | T I | Apache-2.0 | |
| 33 | Qwen3 30B-A3B Thinking 2507Alibaba (Qwen) · CN | 59.36/6 | 80.9 | 73.4 | 85.0 | 66.0 | 22.0* | 30.5B | 3.3B | 262K | — | 2025-07-30 | T | Apache-2.0 | |
| 34 | K2-V2 InstructLLM360 / MBZUAI (K2) · AE · fully open | 57.92/6 | — | 69.3 | 80.2 | — | — | 70B | dense | 512K | — | 2025-12-03 | T | Apache-2.0 | |
| 35 | Qwen3.5 4BAlibaba (Qwen) · CN | 57.83/6 | 79.1 | 76.2 | — | 55.8 | — | 4B | dense | 262K | — | 2026-03-02 | T I V | Apache-2.0 | |
| 36 | MiniCPM5 2BOpenBMB (MiniCPM) · CN | 57.56/6 | 70.8 | 70.2 | 86.5 | 69.1 | 46.4 | 2.5B | dense | 131K | — | 2026-09-06 | T | Apache-2.0 | |
| 37 | Falcon H1R 7BTII (Falcon) · AE · unverified | 57.54/6 | 72.1 | 61.3 | 83.1 | — | — | 7.6B | dense | 262K | — | 2026-01-05 | T | Falcon LLM License | |
| 38 | OLMo 3.1 32B ThinkAi2 (OLMo) · US · fully open | 57.03/6 | — | 57.5 | 78.1 | 83.3 | — | 32.2B | dense | 64K | — | 2025-12-10 | T | Apache-2.0 | |
| 39 | Gemma 4 12BGoogle · US | 56.95/6 | 77.2 | 78.8 | 77.5 | 72.0 | — | 11.9B | dense | 256K | — | 2026-06-03 | T I A | Apache-2.0 | |
| 40 | Magistral Small 1.2Mistral AI · FR · unverified | 55.93/6 | — | 70.1 | 77.3 | 70.9 | — | 24B | dense | 128K | — | 2025-09-17 | T I | Apache-2.0 | |
| 41 | Qwen3 4B Thinking 2507Alibaba (Qwen) · CN · unverified | 55.74/6 | 74.0 | 65.8 | 81.3 | 55.2 | — | 4B | dense | 262K | — | 2025-08-06 | T | Apache-2.0 | |
| 42 | Ministral 3 8BMistral AI · FR | 55.23/6 | — | 66.8 | 78.7 | 61.6 | — | 8B | dense | 256K | $0.15 / $0.15 | 2025-12-02 | T I | Apache-2.0 | |
| 43 | Granite 4.2 3BIBM (Granite) · US | 54.84/6 | 67.8 | 54.8 | 78.3 | 69.7 | — | 3B | dense | 131K | — | 2026-08-07 | T | Apache-2.0 | |
| 44 | MiniCPM-V 4.7 35B-A3BOpenBMB (MiniCPM) · CN | 54.53/6 | 83.5 | — | 76.5 | 56.6 | — | 35.2B | 3B | 262K | — | 2026-10-07 | T I V | Apache-2.0 | |
| 45 | Phi-4-reasoning-plusMicrosoft (Phi) · US · unverified | 54.44/6 | 76.0 | 69.3 | 78.0 | 53.1 | — | 14.7B | dense | 32K | — | 2025-04-30 | T | MIT | |
| 46 | DeepSeek-R1-0528-Qwen3-8BDeepSeek · CN · unverified | 53.13/6 | — | 61.1 | 76.3 | 60.5 | — | 8.2B | dense | 128K | — | 2025-05-29 | T | MIT | |
| 47 | DiffusionGemma 26B-A4BGoogle · US | 52.65/6 | 77.6 | 73.2 | 69.1 | 69.1 | — | 25.8B | 4B | 256K | — | 2026-06-09 | T I | Apache-2.0 | |
| 48 | MiniCPM-SALAOpenBMB (MiniCPM) · CN | 52.63/6 | 67.0 | — | 78.3 | 52.0 | — | 9.5B | dense | 1M | — | 2026-02-11 | T | Apache-2.0 | |
| 49 | Nemotron Nano 9B v2NVIDIA · US · unverified | 52.53/6 | — | 64.0 | 72.1 | 71.1 | — | 8.9B | dense | 128K | — | 2025-08-18 | T | NVIDIA Open Model | |
| 50 | Hunyuan 7B InstructTencent (Hunyuan) · CN · unverified | 52.53/6 | — | 60.1 | 75.3 | 57.0 | — | 7.5B | dense | 256K | — | 2025-08-04 | T | Tencent Hunyuan Community | |
| 51 | Qwen3 32BAlibaba (Qwen) · CN · unverified | 52.42/6 | — | — | 72.9 | 65.7 | — | 32.8B | dense | 131K | $0.080 / $0.28 | 2025-04-29 | T | Apache-2.0 | |
| 52 | Ling 3.0 TinyInclusionAI (Ling) · CN · unverified | 51.93/6 | 65.8* | 73.4 | 73.3* | — | — | 7.9B | 1.4B | 131K | — | 2026-08-10 | T | MIT | |
| 53 | Hermes 4.3 36BNous Research (Hermes) · US | 51.43/6 | 80.7 | 65.5 | 69.3 | — | — | 36.2B | dense | 512K | — | 2025-11-17 | T | Apache-2.0 | |
| 54 | DeepSeek-R1-Distill-Qwen-32BDeepSeek · CN · unverified | 50.93/6 | — | 62.1 | 72.6 | 57.2 | — | 32.8B | dense | 128K | — | 2025-01-20 | T | MIT | |
| 55 | Kanana-2 30B-A3B ThinkingKakao (Kanana) · KR | 50.74/6 | 74.2 | 57.8 | 74.0 | 58.8 | — | 30.7B | 3B | 32K | — | 2026-01-14 | T | Kanana License | |
| 56 | Spark-X2.6 1.7BSparkLLM (XHToken) · CN | 50.14/6 | — | 56.1 | 70.1 | — | 59.8 | 1.7B | dense | 1M | — | 2026-10-19 | T | Apache-2.0 | |
| 57 | DeepSeek-R1-Distill-Llama-70BDeepSeek · CN · unverified | 50.13/6 | — | 65.2 | 70.0 | 57.5 | — | 70.6B | dense | 128K | — | 2025-01-20 | T | MIT | |
| 58 | Ministral 3 3BMistral AI · FR | 49.43/6 | — | 53.4 | 72.1 | 54.8 | — | 3B | dense | 256K | $0.10 / $0.10 | 2025-12-02 | T I | Apache-2.0 | |
| 59 | Qwen3 8BAlibaba (Qwen) · CN · unverified | 48.53/6 | — | 62.0 | 67.3 | 57.5 | — | 8.2B | dense | 131K | — | 2025-04-29 | T | Apache-2.0 | |
| 60 | Edge0 8B-A1B (preview)Edge0 · ? | 47.23/6 | 70.1 | 70.7 | 63.3 | — | — | 7.9B | 1.4B | 128K | — | 2026-09-08 | T | Apache-2.0 | |
| 61 | Phi-4Microsoft (Phi) · US · unverified | 44.72/6 | 70.4 | 56.1 | — | — | — | 14.7B | dense | 16K | $0.056 / $0.11 | 2024-12-12 | T | MIT | |
| 62 | Tri-21B ThinkTrillion Labs · KR | 43.65/6 | 74.3 | 62.6 | 56.7 | 53.7 | — | 20.7B | dense | 32K | — | 2026-02-18 | T | Apache-2.0 | |
| 63 | OLMo 3 7B ThinkAi2 (OLMo) · US · fully open | 43.42/6 | — | 46.2 | 64.6 | — | — | 7.3B | dense | 64K | — | 2025-11-20 | T | Apache-2.0 | |
| 64 | LFM2.5 2.6BLiquid AI · US | 43.25/6 | — | 57.8 | 51.9 | 59.4 | 18.0 | 2.7B | dense | 128K | $0.35 / $0.50 | 2026-07-28 | T | LFM Open License | |
| 65 | Llama 3.3 70BMeta · US · unverified | 40.82/6 | 68.9 | 50.5 | — | — | — | 70.6B | dense | 128K | $0.10 / $0.32 | 2024-12-06 | T | Llama 3.3 Community | |
| 66 | MiMo-7B-RLXiaomi (MiMo) · CN · unverified | 40.74/6 | 58.6 | 54.4 | 55.4 | 49.3 | — | 7.8B | dense | 32K | — | 2025-04-30 | T | MIT | |
| 67 | DeepSeek-R1-Distill-Qwen-7BDeepSeek · CN · unverified | 39.83/6 | — | 49.1 | 55.5 | 37.6 | — | 7.6B | dense | 128K | — | 2025-01-20 | T | MIT | |
| 68 | Mistral Small 3.2Mistral AI · FR · unverified | 38.62/6 | 69.1 | 46.1 | — | — | — | 24B | dense | 128K | $0.075 / $0.20 | 2025-06-20 | T I | Apache-2.0 | |
| 69 | Gemma 4 E4BGoogle · US | 38.24/6 | 69.4 | 58.6 | 42.5 | 52.0 | — | 8B | dense | 128K | — | 2026-04-02 | T I A | Apache-2.0 | |
| 70 | EXAONE 4.0 1.2BLG AI Research (EXAONE) · KR | 36.14/6 | 59.3 | 52.0 | 45.2 | 45.3 | — | 1.3B | dense | 64K | — | 2025-07-15 | T | EXAONE License (non-commercial) | |
| 71 | Qwen3.5 2BAlibaba (Qwen) · CN | 33.02/6 | 66.5 | 51.6 | — | — | — | 2.3B | dense | 262K | — | 2026-03-02 | T I V | Apache-2.0 | |
| 72 | Gemma 3 27BGoogle · US · unverified | 32.73/6 | 67.5 | 42.4 | — | 29.7 | — | 27.4B | dense | 128K | $0.080 / $0.16 | 2025-03-12 | T I | Gemma | |
| 73 | Gemma 4 E2BGoogle · US | 31.94/6 | 60.0 | 43.4 | 37.5 | 44.0 | — | 5.1B | dense | 128K | — | 2026-04-02 | T I A | Apache-2.0 | |
| 74 | SmolLM3 3BHugging Face (SmolLM) · FR · fully open | 30.43/6 | — | 41.7 | 36.7 | 30.0 | — | 3.1B | dense | 64K | — | 2025-07-08 | T | Apache-2.0 | |
| 75 | LFM2.5 1.2B ThinkingLiquid AI · US | 30.13/6 | 49.7 | 37.9 | 31.7 | — | — | 1.2B | dense | 32K | — | 2026-01-20 | T | LFM Open License | |
| 76 | LLaDA2.2 miniInclusionAI (Ling) · CN | 29.73/6 | — | 44.4 | 35.1 | 28.1 | — | 16.3B | 1.4B | 131K | — | 2026-09-05 | T | Apache-2.0 | |
| 77 | Gemma 3 12BGoogle · US · unverified | 28.43/6 | 60.6 | 40.9 | — | 24.6 | — | 12.2B | dense | 128K | $0.040 / $0.12 | 2025-03-12 | T I | Gemma | |
| 78 | Llama 3.1 8BMeta · US · unverified | 21.32/6 | 48.3 | 30.4 | — | — | — | 8B | dense | 128K | $0.020 / $0.040 | 2024-07-23 | T | Llama 3.1 Community | |
| 79 | LFM2.5 1.2B InstructLiquid AI · US | 20.63/6 | 44.4 | 38.9 | 14.0 | — | — | 1.2B | dense | 32K | — | 2026-01-06 | T | LFM Open License | |
| 80 | Apertus 1.5 8BSwiss AI Initiative (Apertus) · CH · fully open · unverified | 20.12/6 | 48.8 | 27.2 | — | — | — | 8.9B | dense | 262K | — | 2026-07-24 | T I A | Apache-2.0 | |
| 81 | EuroLLM 22B InstructEuroLLM (UTTER consortium) · PT · fully open | 19.22/6 | 50.9 | 26.8 | — | — | — | 22.6B | dense | 32K | — | 2025-12-05 | T | Apache-2.0 | |
| 82 | Gemma 3n E4BGoogle · US | 18.24/6 | 50.6 | 23.7 | 11.6 | 25.7 | — | 7.8B | dense | 32K | — | 2025-06-26 | T I A V | Gemma | |
| 83 | Gemma 3 4BGoogle · US · unverified | 17.53/6 | 43.6 | 30.8 | — | 12.6 | — | 4.3B | dense | 128K | $0.040 / $0.050 | 2025-03-12 | T I | Gemma | |
| 84 | Gemma 3n E2BGoogle · US | 14.34/6 | 40.5 | 24.8 | 6.7 | 18.6 | — | 5.4B | dense | 32K | — | 2025-06-26 | T I A V | Gemma | |
| 85 | LFM2.5 350MLiquid AI · US | 12.62/6 | 20.0 | 30.6 | — | — | — | 350M | dense | 32K | — | 2026-03-31 | T | LFM Open License | |
| 86 | Qwen3.5 0.8BAlibaba (Qwen) · CN | 10.72/6 | 42.3 | 11.9 | — | — | — | 900M | dense | 262K | — | 2026-03-02 | T I V | Apache-2.0 | |
| 87 | Gemma 3 1BGoogle · US · unverified | 6.13/6 | 14.7 | 19.2 | — | 1.9 | — | 1B | dense | 32K | — | 2025-03-12 | T | Gemma | |
| 88 | LFM3-Omni 2BLiquid AI · US | — | — | — | — | — | — | 2.2B | dense | 128K | — | 2026-10-21 | T I A V | LFM Open License | |
| 89 | LFM3-VL 2BLiquid AI · US | — | — | — | — | — | — | 1.9B | dense | 128K | — | 2026-10-19 | T I | LFM Open License | |
| 90 | Sage-A1.5Shanghai AI Laboratory (InternScience) · CN | — | — | — | — | — | — | 36B | 3B | 262K | — | 2026-10-15 | T I | Apache-2.0 | |
| 91 | MiniCPM-V 4.7 1BOpenBMB (MiniCPM) · CN | — | — | — | — | — | — | 1.6B | dense | 262K | — | 2026-10-06 | T I V | Apache-2.0 | |
| 92 | Nemotron 3.5 Bolt E4BNVIDIA · US | — | — | — | — | — | — | 8B | dense | 65K | — | 2026-10-05 | T I A | OpenMDW-1.1 | |
| 93 | FrogNano 4BMicrosoft (Phi) · US | — | — | — | — | — | — | 4.7B | dense | 262K | — | 2026-09-17 | T | MIT | |
| 94 | Luciole 23B ThinkingOpenLLM-France (Luciole) · FR | — | — | — | — | — | — | 23.2B | dense | 131K | — | 2026-09-09 | T | Apache-2.0 | |
| 95 | MiMo-V2.6-Distill-Qwen-9BXiaomi (MiMo) · CN · unverified | — | — | — | — | — | — | 9.4B | dense | 262K | — | 2026-09 | T | MIT | |
| 96 | LFM2.5-VL 3BLiquid AI · US | — | — | — | — | — | — | 3.1B | dense | 32K | — | 2026-08-11 | T I | LFM Open License | |
| 97 | Apertus 1.5 70BSwiss AI Initiative (Apertus) · CH · fully open · unverified | — | — | — | — | — | — | 72B | dense | 262K | — | 2026-07-24 | T I A | Apache-2.0 | |
| 98 | Luciole 8B InstructOpenLLM-France (Luciole) · FR | — | — | — | — | — | — | 8.1B | dense | 131K | — | 2026-07-07 | T | Apache-2.0 | |
| 99 | ALIA 40B InstructBarcelona Supercomputing Center (ALIA) · ES | — | — | — | — | — | — | 40.4B | dense | 164K | — | 2026-06-30 | T | Apache-2.0 | |
| 100 | Salamandra 7B InstructBarcelona Supercomputing Center (ALIA) · ES | — | — | — | — | — | — | 7.8B | dense | 164K | — | 2026-06-30 | T | Apache-2.0 | |
| 101 | North Mini CodeCohere · CA · unverified | — | — | — | — | — | — | 30B | 3B | 256K | — | 2026-06-17 | T | Unverified | |
| 102 | LFM2.5 8B-A1BLiquid AI · US | — | — | — | 42.5 | — | — | 8.5B | 1B | 128K | — | 2026-05-28 | T | LFM Open License | |
| 103 | MiniCPM5 1BOpenBMB (MiniCPM) · CN | — | — | — | — | — | — | 1.1B | dense | 131K | — | 2026-05-21 | T | Apache-2.0 | |
| 104 | MiniCPM-V 4.6 ThinkingOpenBMB (MiniCPM) · CN | — | — | — | — | — | — | 1.3B | dense | 262K | — | 2026-05-08 | T I V | Apache-2.0 | |
| 105 | Nemotron 3 Nano Omni 30B-A3BNVIDIA · US · unverified | — | — | — | — | — | — | 31.6B | 3.5B | 256K | — | 2026-04-28 | T I A V | NVIDIA Open Model | |
| 106 | Laguna XS.2poolside · US | — | — | — | — | — | 69.9 | 33B | 3B | 262K | — | 2026-04-28 | T | Apache-2.0 | |
| 107 | Rnj-1.5 InstructEssential AI (Rnj) · US | — | — | — | — | — | — | 8.3B | dense | 164K | — | 2026-04-22 | T | Apache-2.0 | |
| 108 | MiniCPM-V 4.6OpenBMB (MiniCPM) · CN | — | — | — | — | — | — | 1.3B | dense | 262K | — | 2026-04-13 | T I V | Apache-2.0 | |
| 109 | Phi-4-reasoning-vision-15BMicrosoft (Phi) · US · unverified | — | — | — | — | — | — | 15B | dense | 32K | — | 2026-03 | T I | MIT | |
| 110 | TildeOpen 30BTilde · LV | — | — | — | — | — | — | 30.7B | dense | 64K | — | 2026-02-04 | T | CC-BY-4.0 | |
| 111 | MiniCPM-o 4.5OpenBMB (MiniCPM) · CN | — | — | — | — | — | — | 9.4B | dense | 41K | — | 2026-02-03 | T I A V | Apache-2.0 | |
| 112 | Qwen3-Coder-NextAlibaba (Qwen) · CN · unverified | — | — | — | — | — | 70.6 | 79.7B | 3B | 262K | $0.12 / $0.80 | 2026-01-30 | T | Apache-2.0 | |
| 113 | EuroLLM 9B Instruct (2512)EuroLLM (UTTER consortium) · PT · fully open | — | — | — | — | — | — | 9.2B | dense | 32K | — | 2026-01-26 | T | Apache-2.0 | |
| 114 | AgentCPM-Explore 4BOpenBMB (MiniCPM) · CN | — | — | — | — | — | — | 4B | dense | 262K | — | 2026-01-11 | T | Apache-2.0 | |
| 115 | Step3-VL-10BStepFun · CN · unverified | — | — | — | — | — | — | 10B | dense | 64K | — | 2026-01 | T I | Apache-2.0 | |
| 116 | Nemotron 3 Nano 4BNVIDIA · US · unverified | — | — | — | — | — | — | 4B | dense | 262K | — | 2026 | T | NVIDIA Open Model | |
| 117 | HyperCLOVA X SEED Omni 8BNAVER (HyperCLOVA X) · KR · unverified | — | — | — | — | — | — | 10.7B | dense | 8K | — | 2025-12-23 | T I A | HyperCLOVA X SEED License | |
| 118 | Devstral Small 2Mistral AI · FR | — | — | — | — | — | 68.0 | 24B | dense | 256K | — | 2025-12-09 | T I | Apache-2.0 | |
| 119 | GLM-4.6V-FlashZ.ai (GLM) · CN · unverified | — | — | — | — | — | — | 9B | dense | 128K | — | 2025-12-08 | T I | MIT | |
| 120 | Trinity MiniArcee AI (Trinity) · US · unverified | — | — | 58.6 | — | — | — | 26.1B | 3B | 131K | — | 2025-12-01 | T | OpenMDW-1.1 | |
| 121 | BaguettotronPleIAs · FR · fully open | — | — | — | — | — | — | 320M | dense | 4K | — | 2025-11-10 | T | Apache-2.0 | |
| 122 | Bielik 11B v3 InstructSpeakLeash (Bielik) · PL · unverified | — | — | — | — | — | — | 11.2B | dense | 32K | — | 2025-11-07 | T | Apache-2.0 | |
| 123 | ERNIE 4.5 VL 28B-A3B ThinkingBaidu (ERNIE) · CN · unverified | — | — | — | — | — | — | 28B | 3B | 128K | — | 2025-11-07 | T I V | Apache-2.0 | |
| 124 | Kimi Linear 48B-A3BMoonshot AI (Kimi) · CN · unverified | — | — | — | — | — | — | 48B | 3B | 1M | — | 2025-10-30 | T | MIT | |
| 125 | gpt-oss-safeguard-20bOpenAI · US | — | — | — | — | — | — | 20.9B | 3.6B | 131K | — | 2025-10-29 | T | Apache-2.0 | |
| 126 | Ring-mini-2.0InclusionAI (Ling) · CN · unverified | — | — | — | — | — | — | 16.3B | 1.4B | 128K | — | 2025-09-19 | T | MIT | |
| 127 | Ling-mini-2.0InclusionAI (Ling) · CN · unverified | — | — | — | — | — | — | 16.3B | 1.4B | 128K | — | 2025-09-09 | T | MIT | |
| 128 | ERNIE 4.5 21B-A3B ThinkingBaidu (ERNIE) · CN · unverified | — | — | — | — | — | — | 21.8B | 3B | 128K | — | 2025-09-08 | T | Apache-2.0 | |
| 129 | Gemma 3 270MGoogle · US | — | — | — | — | — | — | 270M | dense | 32K | — | 2025-08-14 | T | Gemma | |
| 130 | Qwen3-Coder 30B-A3BAlibaba (Qwen) · CN · unverified | — | — | — | — | — | 51.6 | 30.5B | 3.3B | 262K | $0.15 / $0.60 | 2025-07-31 | T | Apache-2.0 | |
| 131 | Teuken 7B Instruct v0.6OpenGPT-X (Teuken) · DE | — | — | — | — | — | — | 7.5B | dense | 4K | — | 2025-07-28 | T | CC-BY-NC-4.0 | |
| 132 | Voxtral Small 24BMistral AI · FR · unverified | — | — | — | — | — | — | 24B | dense | 32K | $0.10 / $0.30 | 2025-07-15 | T A | Apache-2.0 | |
| 133 | ERNIE 4.5 0.3BBaidu (ERNIE) · CN | — | — | — | — | — | — | 360M | dense | 128K | — | 2025-06-30 | T | Apache-2.0 | |
| 134 | Kimi-VL-A3B-Thinking-2506Moonshot AI (Kimi) · CN · unverified | — | — | — | — | — | — | 16.4B | 2.8B | 128K | — | 2025-06-21 | T I V | MIT | |
| 135 | MiMo-VL-7B-RLXiaomi (MiMo) · CN · unverified | — | — | — | — | — | — | 8.3B | dense | 32K | — | 2025-05-30 | T I V | MIT | |
| 136 | Poro 2 70B InstructLumiOpen / AMD Silo AI (Poro) · FI | — | — | — | — | — | — | 70.6B | dense | 8K | — | 2025-05-27 | T | Llama 3.3 Community | |
| 137 | Phi-4-miniMicrosoft (Phi) · US · unverified | — | 52.8 | — | — | — | — | 3.8B | dense | 128K | — | 2025-02-26 | T | MIT | |
| 138 | Llama 3.2 3BMeta · US · unverified | — | — | 32.8 | — | — | — | 3.2B | dense | 128K | $0.050 / $0.33 | 2024-09-25 | T | Llama 3.2 Community | |
| 139 | Llama 3.2 1BMeta · US · unverified | — | — | 27.2 | — | — | — | 1.2B | dense | 128K | $0.027 / $0.20 | 2024-09-25 | T | Llama 3.2 Community |
Scores are self-reported by each lab unless marked: ✓ independently measured, * measured by another lab. "—" means not reported. Hover a score for its source and variant (e.g. AIME year). Input: T text · I image · A audio · V video.
The leaderboard is sorted by the Small LM Index: the expected average on six core benchmarks, with any the lab didn't report estimated from its other scores (how it works). Tick up to four models to compare them.