Skip to content

Meta vs Google vs DeepSeek vs OpenAI

Each lab shows its best-ranked model on every board. One row per board where at least one is ranked, from each board's current capture.

  • Meta
  • Google
  • DeepSeek
  • OpenAI

Head to head on 26 boards: OpenAI places best on 14, Google places best on 6, Meta places best on 5 and DeepSeek places best on 1. All 4 ranked on 15 of 28 boards.

Two at a time on this screen. Tap one to swap it in.

BoardMetaGoogleDeepSeekOpenAI
Chat2
MultiChallengeScoretoday#175.5%best of the pickedMuse Spark#371.4%Gemini 3.1 Pro Preview#2446.1%DeepSeek V3P1#469.2%GPT-5.4 Pro
Text ArenaArena ratingtoday#61,496best of the pickedMuse Spark 1.2#81,492Gemini 3.8 Flash#251,477DeepSeek V4.1 Flash#161,483GPT-5.6 Sol
Reasoning7
GPQA DiamondAccuracytoday#3589.8%Muse Spark#295.4%Gemini 3.8 Flash#2091.7%DeepSeek V4 Pro 0813#195.8%best of the pickedGPT-6 Astra
Epoch ECIECItoday#15156.9Muse Spark 1.3#11157.7Gemini 3.7 Flash#24155.4DeepSeek V4 Pro 0813#1166.6best of the pickedGPT-6 Astra
ARC-AGI-2Scoretoday#710.0%Llama 4 Maverick Instruct#689.2%Gemini 3.8 Flash#2361.4%DeepSeek V4 Flash Vision#195.0%best of the pickedGPT-6 Astra
MMLU-ProAccuracy · staletoday#4782.8%Llama4 Behemoth#191.2%best of the pickedGemini 3.1 Pro Preview#3385.0%DeepSeek-V3.2#389.3%o1
AA IntelligenceIndextoday#448.1Muse Spark 1.3#1540.9Gemini 3.8 Flash#1839.5DeepSeek V4.1 Flash#352.7best of the pickedGPT-6 Astra
LiveBenchGlobal averagetoday#581.6%Muse Spark 1.3#1378.8%Gemini 3.7 Flash#681.1%DeepSeek V4.1 Flash#482.2%best of the pickedGPT-6 Astra
HLEAccuracytoday#640.6%Muse Spark#346.4%Gemini 3.1 Pro Preview#154.8%best of the pickedGPT-6 Astra
Coding4
Aider PolyglotPass rate · staletoday#4315.6%Llama 4 Maverick Instruct#383.1%Gemini 2.5 Pro (Jun 2025)#874.2%DeepSeek-V3.2#188.0%best of the pickedGPT-5
WebDev ArenaArena ratingtoday#81,656Muse Spark 1.3#201,593Gemini 3.7 Flash#131,621DeepSeek V4.1 Flash#21,792best of the pickedGPT-6 Astra
SWE-bench VerifiedResolved · staletoday#3621.0%Llama 4 Maverick Instruct#275.8%best of the pickedGemini 3 Flash#1170.0%DeepSeek-V3.2#672.8%GPT-5.2
SWE-Bench ProResolvedtoday#958.8%Gemini 3.8 Flash#390.2%best of the pickedGPT-6 Astra
Agents and tools3
BFCLOverall accuracytoday#3637.3%Llama 4 Maverick 17b 128e#372.5%best of the pickedGemini 3 Pro#1256.7%DeepSeek-V3.2#763.0%o3
MCP AtlasPass ratetoday#188.1%best of the pickedMuse Spark 1.1#683.6%Gemini 3.5 Flash#1181.8%GPT-5.6
Search ArenaArena ratingtoday#91,210Gemini 3.1 Pro Grounding#11,257best of the pickedGPT-5.6 Sol
Vision2
Vision ArenaArena ratingtoday#51,294best of the pickedMuse Spark 1.3#101,289Gemini 3 Pro#111,287GPT-5.5
Document ArenaArena ratingtoday#121,471Muse Spark 1.3#161,463Gemini 3.5 Flash#61,486best of the pickedGPT-5.5
Image2
Text-to-ImageArena ratingtoday#71,276Muse Image#91,261Gemini 3.1 Flash Image#11,424best of the pickedGPT Image 2.5 Sunburst
Image EditArena ratingtoday#61,402Muse Image#91,390Gemini 3 Pro Image#11,526best of the pickedGPT Image 2.5 Sunburst
Video2
Text-to-VideoArena ratingtoday#91,456Muse Video#11,516best of the pickedGemini Omni 1.1 Flash#111,368Sora 2 Pro
Image-to-VideoArena ratingtoday#21,488Gemini Omni 1.1 Flash
Speech2
Speech ArenaArena ratingtoday#89847Metavoice V1#21,268best of the pickedGemini 3.8 Flash TTS#321,104TTS 1 Hd
Open ASRAverage WERtoday#284.84 WERbest of the pickedmuse-voice-transcribe (meta)#669.51 WERgemma-4-E4B-it (google)#535.78 WERwhisper-large-v3 (openai)
Embeddings1
MTEBMean task scoretoday#568.4%gemini-embedding-001 (google)
Speed1
AA SpeedOutput speedtoday#2232 tok/sMuse Spark 1.3#1308 tok/sbest of the pickedGemini 3.8 Flash#3232 tok/sDeepSeek V4.1 Flash#5154 tok/sGPT-6 Luna
Price1
AA PriceBlended pricetoday#11$2.00Muse Spark 1.3#9$1.50Gemini 3.8 Flash#4$0.53DeepSeek V4.1 Flash#1$0.20best of the pickedGPT-6 Luna
Usage1
OpenRouterWeekly tokenstoday#151.8T tokMuse Spark 1.3 Contributor#142.2T tokGemini 3.8 Flash#118.6T tokbest of the pickedDeepSeek V4.1 Flash#58.6T tokGPT-5.6 Luna

Green marks the best placing among the picked where two or more are ranked; ties are marked for each. Ranks are each board's own order (lower is better), scores as the board published them. Dash: not on that board.

More comparisons

New #1s by email

Saturdays, only in weeks when a leaderboard has a new #1.

Double opt-in. Unsubscribe any time.