A Apodex Anthropic Meta DeepSeek Head to head on 20 boards: Anthropic places best on 15 , Meta places best on 4 , DeepSeek places best on 2 and Apodex places best on 0 (a tie counts for each). All 4 ranked on 2 of 26 boards.
Board Read A Apodex Anthropic Meta DeepSeek Chat2 Text Arena Arena rating 1d yesterday Nov 11 1d yesterday #4 1,506 Claude Opus 5.5 #3 1,511 best of the picked Muse Wonder #32 1,475 DeepSeek V4.1 Flash MultiChallenge Score 1d yesterday Nov 11 1d yesterday #10 67.7% Claude Haiku 5.5 #3 75.5% best of the picked Muse Spark #5 74.9% DeepSeek V4P1 Flash Reasoning7 AA Intelligence Index 1d yesterday Nov 11 1d yesterday #37 13.7 Apodex 1.1 #1 64.6 best of the picked Claude Fable 5.2 #9 50.1 Muse Wonder Full #17 32.6 DeepSeek V4.1 Flash Epoch ECI ECI 1d yesterday Nov 11 1d yesterday #1 174.2 best of the picked Claude Fable 5.2 #15 162.0 Muse Wonder #21 159.3 DeepSeek V4.1 Flash LiveBench Global average 1d yesterday Nov 11 1d yesterday #1 83.4% best of the picked Claude Fable 5.1 #6 81.6% Muse Spark 1.3 #8 81.1% DeepSeek V4.1 Flash GPQA Diamond Accuracy 1d yesterday Nov 11 1d yesterday #4 95.6% best of the picked Claude Sonnet 5.5 #49 89.8% Muse Spark #23 93.2% DeepSeek V4.1 Flash ARC-AGI-2 Score 1d yesterday Nov 11 1d yesterday #4 93.8% best of the picked Claude Fable 5.2 #41 57.5% Muse Spark Open #21 78.1% DeepSeek V4.1 Flash MMLU-Pro Accuracy · stale 1d yesterday Nov 11 1d yesterday #4 89.1% best of the picked Claude Opus 4.6 #47 82.8% Llama4 Behemoth #33 85.0% DeepSeek-V3.2 HLE Accuracy 1d yesterday Nov 11 1d yesterday #2 46.5% best of the picked Claude Fable 5.1 #7 40.6% Muse Spark Coding4 WebDev Arena Arena rating 1d yesterday Nov 11 1d yesterday #1 1,836 best of the picked Claude Fable 5.2 #8 1,723 Muse Wonder Full #15 1,664 DeepSeek V4.1 Flash SWE-bench Verified Resolved · stale 1d yesterday Nov 11 1d yesterday #1 76.8% best of the picked Claude Opus 4.5 #36 21.0% Llama 4 Maverick Instruct #11 70.0% DeepSeek-V3.2 Aider Polyglot Pass rate · stale 1d yesterday Nov 11 1d yesterday #10 72.0% Claude Opus 4 #43 15.6% Llama 4 Maverick Instruct #8 74.2% best of the picked DeepSeek-V3.2 SWE-Bench Pro Resolved 1d yesterday Nov 11 1d yesterday #1 98.0% best of the picked Claude Opus 5 #4 90.2% Muse Wonder 1.0 Max Effort Agents and tools3 BFCL Overall accuracy · stale 1d yesterday Nov 11 1d yesterday #1 77.5% best of the picked Claude Opus 4.5 #36 37.3% Llama 4 Maverick 17b 128e #12 56.7% DeepSeek-V3.2 MCP Atlas Pass rate 1d yesterday Nov 11 1d yesterday #3 87.2% Claude Fable 5.1 #2 88.1% best of the picked Muse Spark 1.1 Search Arena Arena rating 1d yesterday Nov 11 1d yesterday #5 1,229 Claude Fable 5 Vision2 Vision Arena Arena rating 1d yesterday Nov 11 1d yesterday #1 1,309 best of the picked Claude Fable 5 #6 1,294 Muse Spark #35 1,263 DeepSeek V4.1 Flash Document Arena Arena rating 1d yesterday Nov 11 1d yesterday #1 1,500 best of the picked Claude Fable 5 #9 1,480 Muse Spark 1.1 Image2 Text-to-Image Arena rating 1d yesterday Nov 11 1d yesterday #11 1,266 Muse Image Image Edit Arena rating 1d yesterday Nov 11 1d yesterday #12 1,400 Muse Image Video1 Text-to-Video Arena rating 1d yesterday Nov 11 1d yesterday #10 1,457 Muse Video Speech2 Open ASR Average WER 1d yesterday Nov 11 1d yesterday #30 7.00 WER muse-voice-transcribe (meta) Speech Arena Arena rating 1d yesterday Nov 11 1d yesterday #99 846 Metavoice V1 Speed1 AA Speed Output speed 1d yesterday Nov 11 1d yesterday #24 14 tok/s Apodex 1.1 #4 44 tok/s best of the picked Claude Fable 5.2 #25 13 tok/s Muse Wonder #15 24 tok/s DeepSeek V4 Pro 0813 Price1 AA Price Blended price 1d yesterday Nov 11 1d yesterday #1 $0.20 best of the picked Claude Haiku 5.5 #1 $0.20 best of the picked Muse Ember #4 $0.53 DeepSeek V4.1 Flash Usage1 OpenRouter Weekly tokens 1d yesterday Nov 11 1d yesterday #6 6.8T tok Claude Opus 5.5 #15 2.9T tok Muse Spark 1.3 Contributor #1 53.8T tok best of the picked DeepSeek 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.