// ranked · September 2026

Best Free LLM APIs

Models with a free listing on OpenRouter, ranked by LiveBench overall score of the underlying model where one is published. Free tiers carry tighter rate limits than paid ones.

OpenRouter + LiveBenchAll rankingsAll models

Qwen3.8 27B leads on LiveBench overall at 75.3, ahead of GLM 5.2 (73.2) and Inkling (71.9).

Qwen3.8 27B vs GLM 5.2, head to head

Top 10

#ModelOverall scoreBlended / 1MContextCapabilities
1Qwen3.8 27B

qwen/qwen3.8-27b:free

75.3Free1M
reasoningtoolsvision
2GLM 5.2

z-ai/glm-5.2:free

73.2Free1.0M
reasoningtools
3Inkling

thinkingmachines/inkling:free

71.9Free1.0M
reasoningtoolsvision
4Nemotron 3 Ultra

nvidia/nemotron-3-ultra-550b-a55b:free

67.4Free262K
reasoningtools
5Inkling Small

thinkingmachines/inkling-small:free

Not evaluatedFree1.0M
reasoningtoolsvision
6Laguna S 2.1

poolside/laguna-s-2.1:free

Not evaluatedFree1.0M
reasoningtools
7Lyria 3 Pro Preview

google/lyria-3-pro-preview

Not evaluatedFree1.0M
vision
8Lyria 3 Clip Preview

google/lyria-3-clip-preview

Not evaluatedFree1.0M
vision
9Dots3-Note Preview (free)

dots-studio/dots-3-note-preview:free

Not evaluatedFree512K
reasoningtoolsvision
10Nex-N2.5-Mini

nex-agi/nex-n2.5-mini:free

Not evaluatedFree262K
reasoningvision

13 more models qualify — see the full leaderboard.

FAQ

What is the best free LLM API in September 2026?

Qwen3.8 27B leads on LiveBench overall at 75.3, ahead of GLM 5.2 (73.2) and Inkling (71.9).

How is this ranking produced?

Models with a free listing on OpenRouter, ranked by LiveBench overall score of the underlying model where one is published. Free tiers carry tighter rate limits than paid ones.

Other rankings

How this ranking is produced

  • One metric, stated above — nothing here is weighted or scored by us. Scores come from LiveBench release 2026-06-25; models without a published run don't appear in score-based rankings.
  • Price, context and capabilities — live from OpenRouter, refreshed every 15 minutes.
  • Ties — scores less than a point apart are called a tie; effort settings alone move a LiveBench score by more than that.

A public benchmark is someone else's workload. Before committing, see LLM & agent evaluation.