Skip to content

Rankings

Output tokens per second

Performance measure
#ModelTokens per secondAgainst the leaderProvider
1grok-4.7xAI16.5k t/sAtlas Cloud
2Gemini 2.5 ProGoogle7.3k t/sAtlas Cloud
2Gemini 2.5 ProGoogle7.3k t/sDeepInfra
4Gemini 3.1 Flash LiteGoogle5.8k t/sAtlas Cloud
4Gemini 3.1 Flash LiteGoogle5.8k t/sDeepInfra
4Gemini 3.1 Flash LiteGoogle5.8k t/sEachLabs
7Gemini 3.8 FlashGoogle4.5k t/sGoogle
7Gemini 3.8 FlashGoogle4.5k t/sGoogle Vertex AI
9Gemini Pro LatestGoogle4.0k t/sGoogle
10gemini-3.8-flash-developerGoogle3.5k t/sAtlas Cloud
Source 
Hopscotch catalogue
Window 
24 hours
Ranked 
679 of 679 routes
Method
Download

Crowd preference

#ModelRatingAgainst the leaderBoundsVotes
1Gemini 4 ArgonGoogle1,5261,519–1,53210,083
2claude-opus-5-5Anthropic1,5161,509–1,52310,322
3Claude Fable 5.1Anthropic1,5131,507–1,51816,343
4Claude Opus 5Anthropic1,5081,504–1,51334,526
5Claude Opus 4.6Anthropic1,5011,498–1,50583,454
6Gemini 3.8 FlashGoogle1,4981,494–1,50343,997
7Claude Fable 5Anthropic1,4911,487–1,49544,253
7muse-spark-1.3Meta1,4911,486–1,49620,506
7MiMo-V2.6-ProXiaomi1,4911,482–1,5004,067
10Claude Opus 4.7Anthropic1,4901,486–1,49365,140
Snapshot 
text arena, overall, 22 Oct 2026
Matched 
126 of 418 entries
Method
Download

Value for price

#ModelValueLower is betterBlended price / 1MRating
1Llama 3.1 8B InstructMeta0.02$0.02751,187
2GPT OSS 20BOpenAI0.03$0.0361,287
3GLM 5.3 FlashZhipu AI0.04$0.05351,472
4Granite 4.2 3BIBM0.04$0.05251,294
5llama-3.2-1b-instructMeta0.07$0.07051,055
6Microsoft: Phi 4Microsoft0.07$0.08751,217
7GPT OSS 120BOpenAI0.07$0.101,366
8Gemma 4 26B A4BGoogle0.07$0.107251,434
9DeepSeek V4 FlashDeepSeek0.08$0.11251,432
10IBM: Granite 4.2 8BIBM0.08$0.10751,319
Formula 
Blended price ÷ rating × 1,000
Price 
Hopscotch published rates
Rating 
Arena Intelligence, Inc.
Covered 
126 of 126 rated models
Method
Download