Reflection vs Meta vs DeepSeek vs Microsoft

Head to head on 19 boards: DeepSeek places best on 8, Meta places best on 6, Microsoft places best on 4 and Reflection places best on 1. All 4 ranked on 1 of 26 boards.

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

RReflectionMetaDeepSeekMicrosoft

Board	Read	RReflection	Meta	DeepSeek	Microsoft
Chat2
MultiChallengeScoretodaytoday	Oct 21todaytoday		#275.5%best of the pickedMuse Spark	#2646.1%DeepSeek V3P1
Text ArenaArena ratingtodaytoday	Oct 21todaytoday		#81,494best of the pickedMuse Spark 1.3	#341,475DeepSeek V4.1 Flash
Reasoning7
ARC-AGI-2Scoretodaytoday	Oct 21todaytoday	#4623.5%Beam 1.0	#800.0%Llama 4 Maverick Instruct	#2172.9%best of the pickedDeepSeek V4.1 Flash
MMLU-ProAccuracy · staletodaytoday	Oct 21todaytoday		#4782.8%Llama4 Behemoth	#3385.0%best of the pickedDeepSeek-V3.2	#7576.0%Phi 4 Plus
LiveBenchGlobal averagetodaytoday	Oct 21todaytoday		#681.6%best of the pickedMuse Spark 1.3	#881.1%DeepSeek V4.1 Flash
Epoch ECIECItodaytoday	Oct 21todaytoday		#12161.8best of the pickedMuse Wonder	#32155.9DeepSeek V4.1 Flash
GPQA DiamondAccuracytodaytoday	Oct 21todaytoday		#4389.8%Muse Spark	#2591.7%best of the pickedDeepSeek V4 Pro 0813
AA IntelligenceIndextodaytoday	Oct 21todaytoday		#256.0Muse Wonder Full
HLEAccuracytodaytoday	Oct 21todaytoday		#740.6%Muse Spark
Coding4
Aider PolyglotPass rate · staletodaytoday	Oct 21todaytoday		#4315.6%Llama 4 Maverick Instruct	#874.2%best of the pickedDeepSeek-V3.2
WebDev ArenaArena ratingtodaytoday	Oct 21todaytoday		#101,656best of the pickedMuse Spark 1.3	#191,619DeepSeek V4.1 Flash
SWE-bench VerifiedResolved · staletodaytoday	Oct 21todaytoday		#3621.0%Llama 4 Maverick Instruct	#1170.0%best of the pickedDeepSeek-V3.2
SWE-Bench ProResolvedtodaytoday	Oct 21todaytoday		#490.2%Muse Wonder 1.0 Max Effort
Agents and tools2
BFCLOverall accuracy · staletodaytoday	Oct 21todaytoday		#3637.3%Llama 4 Maverick 17b 128e	#1256.7%best of the pickedDeepSeek-V3.2	#5028.8%Phi 4
MCP AtlasPass ratetodaytoday	Oct 21todaytoday		#188.1%Muse Spark 1.1
Vision2
Vision ArenaArena ratingtodaytoday	Oct 21todaytoday		#71,294best of the pickedMuse Spark	#391,260DeepSeek V4.1 Flash
Document ArenaArena ratingtodaytoday	Oct 21todaytoday		#121,471Muse Spark 1.3
Image2
Text-to-ImageArena ratingtodaytoday	Oct 21todaytoday		#81,274Muse Image		#51,332best of the pickedMAI Image 2.6
Image EditArena ratingtodaytoday	Oct 21todaytoday		#71,403Muse Image		#51,428best of the pickedMAI Image 2.6
Video1
Text-to-VideoArena ratingtodaytoday	Oct 21todaytoday		#91,456Muse Video
Speech2
Open ASRAverage WERtodaytoday	Oct 21todaytoday		#356.48 WERmuse-voice-transcribe (meta)		#44.26 WERbest of the pickedazure-speech-06-2026 (microsoft)
Speech ArenaArena ratingtodaytoday	Oct 21todaytoday		#95847Metavoice V1		#291,141best of the pickedAzure Hd 2.5
Embeddings1
MTEBMean task scoretodaytoday	Oct 21todaytoday				#174.3%harrier-oss-v1-27b (microsoft)
Speed1
AA SpeedOutput speedtodaytoday	Oct 21todaytoday	#6345 tok/sbest of the pickedBeam 1.0	#7281 tok/sMuse Spark 1.3	#16209 tok/sDeepSeek V4.1 Flash	#7545 tok/sPhi 4 Mini
Price1
AA PriceBlended pricetodaytoday	Oct 21todaytoday		#15$2.00Muse Spark 1.3	#8$0.53best of the pickedDeepSeek V4.1 Flash
Usage1
OpenRouterWeekly tokenstodaytoday	Oct 21todaytoday		#123.3T tokMuse Spark 1.3 Contributor	#127.3T tokbest of the pickedDeepSeek V4.1 Flash

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.