Compare
Compare models side by side
Pick two to four. We start you with the smartest model, the best open one and the best cheap one. The link in your address bar keeps your picks.
Shapes side by side
Bigger shape, stronger model. A spoke with no data sits at the centre.
Gemini 4 Carbon DeepSeek V4.1 Flash MiMo-V2.6-Pro
Category scores
Percentiles from 0 to 100. The winner in each row is in bold.
Coding
99
88
92
Agents & tools
98
No data yet–
66
Reasoning & knowledge
100
92
98
Math
100
92
96
Chat quality
100
91
97
Following instructions
100
93
97
The numbers behind it
Each tile links to its source. Best value in green. “No data” means the source hasn't tested that model, never a zero.
SIscope scoreHow it works
0 to 100, all categories weighted
Gemini 4 Carbon: 99
DeepSeek V4.1 Flash: 90
MiMo-V2.6-Pro: 88
Arena: textArena ↗
Blind human votes on chat answers
Gemini 4 Carbon: 1540#1
DeepSeek V4.1 Flash: 1465#38
MiMo-V2.6-Pro: 1491#12
Arena: WebDevArena ↗
Human votes on web apps the models build
Gemini 4 Carbon: 1768#4
DeepSeek V4.1 Flash: 1619#29
MiMo-V2.6-Pro: 1625#25
Arena: visionArena ↗
Human votes on questions about images
No dataGemini 4 Carbon: –
DeepSeek V4.1 Flash: 1278#38
MiMo-V2.6-Pro: 1263#53
SWE-bench VerifiedEpoch AI ↗
Fixing real GitHub issues in Python repositories. Epoch AI run, share correct.
No dataGemini 4 Carbon: –
No dataDeepSeek V4.1 Flash: –
No dataMiMo-V2.6-Pro: –
GPQA DiamondEpoch AI ↗
PhD-level science multiple choice questions. Epoch AI run, share correct.
No dataGemini 4 Carbon: –
No dataDeepSeek V4.1 Flash: –
No dataMiMo-V2.6-Pro: –
FrontierMath Tiers 1-3Epoch AI ↗
Very hard, unpublished research-style maths problems. Epoch AI run, share correct.
No dataGemini 4 Carbon: –
No dataDeepSeek V4.1 Flash: –
No dataMiMo-V2.6-Pro: –
OTIS Mock AIME 2024-25Epoch AI ↗
Competition maths problems in the style of AIME. Epoch AI run, share correct.
No dataGemini 4 Carbon: –
No dataDeepSeek V4.1 Flash: –
No dataMiMo-V2.6-Pro: –
SimpleQA VerifiedEpoch AI ↗
Short factual questions; rewards not making things up. Epoch AI run, share correct.
No dataGemini 4 Carbon: –
No dataDeepSeek V4.1 Flash: –
No dataMiMo-V2.6-Pro: –
Blended price
Per 1M tokens, 3 parts input to 1 part output. Lower wins.
No dataGemini 4 Carbon: –
No dataDeepSeek V4.1 Flash: –
MiMo-V2.6-Pro: $0.54
Context window
Most tokens one request can hold
No dataGemini 4 Carbon: –
No dataDeepSeek V4.1 Flash: –
MiMo-V2.6-Pro: 1.1M
Arena scores work like chess ratings; the small number is the rank on that board. Prices are OpenRouter list prices for the exact model. Speed: coming soon.