Home/Open-weights ranking

Open-weights ranking

13 models with published weights. The licence column matters more than the score — two "open" models can have wildly different commercial terms.

Best open result
46.0v5
Kimi K3.1 Pro
Gap to closed
14.0pts
Best closed 60.0
Agent tasks (TB 4.0)
2.0pt
Open in front: open 67.2% vs closed 65.2%
Best single-machine model
33.7Apache-2.0
Qwen3.8 27B
#ModelLicenceParametersContextv4.3.2v5GPQATB 4.0Output $
1 Kimi K3.1 Pro Open weightsMoonshot AI Kimi K3 License 2.8T/104B 1M 55.1 46.0 — 67.2 $10.00
2 MiMo-V2.6 Pro Open weightsXiaomi MIT 1.02T/42B 1M 46.3 30.0 59.5 31.3 $1.40
3 DeepSeek V4.1 Flash Open weightsDeepSeek MIT 552B/16B 1.05M 45.1 33.0 89.8 19.7 $1.20
4 GLM-5.3 Open weightsZ.ai GLM-5.3 License 753B/18B 1M 44.8 32.0 90.9 41.9 $4.40
5 Kimi K3 Open weightsMoonshot AI Kimi K3 License 2.78T/104B 1M 43.6 — 93.1 17.2 $15.00
6 GLM-5.3 Flash Open weightsZ.ai MIT 321B/18B 1M 41.8 — — — $0.50
7 Qwen3.8 27B Open weightsAlibaba Qwen Apache-2.0 27.8B 262K 33.7 — — — $0.40
8 MiniMax M3 Open weightsMiniMax MiniMax Community License 428B/23B 1M 29.2 — 90.9 1.0 $1.20
9 Mistral Large 4 Open weightsMistral Mistral Ecosystem Open-Weight License 1T/52B 1M — 26.0 87.6 — $2.10
10 DeepSeek V4 Pro Open weightsDeepSeek MIT 1.6T/49B 1.05M — 23.0 — 14.1 $3.96
11 Gemma 4 31B Open weightsGoogle Apache-2.0 31B 262K 19.0 — 75.8 — $0.40
12 gpt-oss-120b Open weightsOpenAI Apache-2.0 117B/5.1B 131K — — 75.8 — $0.20
13 Nemotron 3 Ultra Open weightsNVIDIA OpenMDW-1.1 550B/55B 1M — — — — $1.60
Ranking uses the v4.3.2 scale, where open-model coverage is most complete (9 of 13 have results). A separate v5 column is shown for reference; the two are not directly comparable. Models with only a v5 result are ranked on that score, so their position may be slightly understated.

Open vs closed: where the gap actually is

Knowledge tasks are essentially level. On GPQA Diamond the best open model, Kimi K3, scores 93.1% against 96.4% for the best closed model, Gemini 4 Argon — a 3.3-point gap. A year ago that gap was more than 15 points.

The agent ceiling has changed hands. With Kimi K3.1 Pro's weights opened on 11-04, the top of Terminal-Bench 4.0 (67.2%) belongs to an open-weights model for the first time; the best closed model, Gemini 4 Carbon, is at 65.2%. The crown has costs: 2.8T parameters need dozens of GPUs to self-host, and 28 tok/s is the slowest output among leading models — most teams will still use its API.

Licences trip people up more often than scores. MIT and Apache-2.0 are free for commercial use; the Kimi K3 licence requires a visible attribution past 100M monthly users; the MiniMax Community License and OpenMDW-1.1 each carry their own commercial limits. Read the licence before you read the score.

Self-host or call the API

When self-hosting pays

Steady high volume, data that cannot leave the network, and a need to fine-tune. That is where 30B-class models like Qwen3.8-27B or GLM-5.3 Flash hit the sweet spot.

When the API pays

Volatile traffic, a need for frontier capability, or no GPU operations team. Self-hosting a 2.8T model like Kimi K3 would take dozens of H100s — not a serious conversation.

Hybrid is the norm

Simple tasks go to a local small model, hard ones route to an API. With OpenRouter or a self-built gateway in front, cost drops below 30% of an API-only setup.