Open Source LLM Statistics (2026)
As of September 29, 2026, 11 of the top 50 AI models on BenchLM (22%) are open-weight.
Key statistics
- As of September 29, 2026, 11 of the top 50 AI models on BenchLM (22%) are open-weight.
- As of September 29, 2026, the highest-ranked open-weight model on BenchLM is Xiaomi's MiMo-V2.6-Pro at 74.18/100 and #9 overall.
- As of September 29, 2026, 94 open-weight models receive a current overall score in BenchLM's ranked cohort.
- The gap between the top proprietary model (GPT-6 Astra) and the top open-weight model (MiMo-V2.6-Pro) is 14.04 points on BenchLM's 100-point overall score as of September 29, 2026.
- 52% of the 523 AI models tracked by BenchLM (271 models) are open-weight as of September 29, 2026.
- The Arena Elo gap between the top closed and top open model narrowed to 18.4 points by 2026-09-01, from 32.4 points roughly a year earlier, per Arena data tracked on BenchLM.
As of September 29, 2026, the highest-ranked open-weight model on BenchLM is Xiaomi's MiMo-V2.6-Pro at 74.18/100 and #9 overall.
MiMo-V2.6-Pro (74.18/100)
As of September 29, 2026, 94 open-weight models receive a current overall score in BenchLM's ranked cohort.
94 models
The gap between the top proprietary model (GPT-6 Astra) and the top open-weight model (MiMo-V2.6-Pro) is 14.04 points on BenchLM's 100-point overall score as of September 29, 2026.
14.04 points
Methodology & sources
"Open-weight" follows the sourceType field in BenchLM's catalog. Rankings use the active BenchAlign overall score, rebuilt from current inputs on every build. Each position is labeled Supported or Estimated within one shared ordering. Arena Elo figures come from the Arena leaderboard snapshot tracked on BenchLM.
Cite these statistics
Every number on this page is generated from BenchLM's live dataset and refreshed with each data update. Link any statistic directly using its anchor, or cite the page as:
BenchLM.ai, "LLM Statistics" (September 29, 2026), https://benchlm.ai/stats/open-source-llmQuestions
How close are open-source models to closed models?
As of September 29, 2026, the top open-weight model (MiMo-V2.6-Pro) ranks #9 overall on BenchLM, 14.04 points behind the top proprietary model (GPT-6 Astra) on a 100-point scale.
What share of top AI models are open-weight?
22% of BenchLM's overall top 50 (11 models) are open-weight as of September 29, 2026.
What is the best open-source LLM by benchmark score?
As of September 29, 2026, Xiaomi's MiMo-V2.6-Pro is the highest-ranked open-weight model on BenchLM at 74.18/100 and #9 overall. “Open-weight” describes downloadable weights; the page separates OSI-approved licenses from community and custom terms.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 2,000+ readers.
One email each week. Unsubscribe anytime.