Skip to content
Model Pareto

Compare models

Side by side: quality per category, price, speed, and every benchmark cell with where it came from. Pick up to four.
DeepSeek V4.1 FlashGLM-5.3 NVFP4Hy4 preview
Focus

Hard reasoning: quality vs price

Compared models are ringed; the other 190 ranked here are greyed.

Lower price is better. Pareto frontier: Gemma 4 E4B, Sarvam 30B, Qwen3.5 4B, Sarvam 105B, Ling-3.0-flash-VL, MiMo-V2.5, GPT-6 Luna, Qwen3.8 Flash-Next, GLM-5.3 Flash, MiMo-V2.6-Pro, Muse Spark 1.3, Claude Sonnet 5.5, Claude Opus 5.5. 80 models are not plotted: Ternary Bonsai 2 27B, DeepSeek V4.1 Flash NVFP4, GLM-5.3 Flash NVFP4, GLM-5.3 NVFP4, Qwen3.8 Flash-Next NVFP4, Qwen3.8 27B NVFP4, Motif 3, Nex-N2-Pro, Ternary Bonsai 27B, K2 Horizon 375B A23B, A.X-K2, Solar Open2 250B, Bonsai 27B, Xing4.0-29B-A4B, K2-Horizon-32B, K2 Horizon MoVA 36B A4B, Maple-Preview, Intern-S2-Preview (35B-A3B), K-EXAONE 2.0 0803, G9v3-39A5B, K2 Horizon 7B, Phi-4-reasoning-plus, EXAONE 4.5 33B, Nanbeige4.1-3B, HyperNova 60B 2605, Nemotron Cascade 2 30B A3B, INTELLECT-3, Apriel-v1.6-15B-Thinker, Command A+, K2 Horizon 3.7B, Ling 3.0 Tiny, K2 Think V2, Step3 VL 10B, DiffusionGemma 26B A4B, North Mini Code, K2-V2, Falcon-H1R-7B, Solar Open 100B, Llama 3.1 Nemotron Ultra 253B v1, MiniCPM5-2B, LongCat Flash Lite, HyperCLOVA X SEED Think (32B), Tri-21B-Think, EXAONE 4.0 32B, Olmo 3.1 32B Think, LFM2.5-2.6B, Devstral 2, LFM2.5-8B-A1B, Olmo 3.1 32B Instruct, Olmo 3 7B Think, NVIDIA Nemotron 3 Nano 4B, Devstral Small 2, Phi-4-mini-flash-reasoning, Qwen3.5 2B, Falcon-H1-34B-Instruct, Hermes 4 - Llama-3.1 70B, LFM2 24B A2B, Exaone 4.0 1.2B, Llama 3.2 Instruct 90B (Vision), Gemma 4 E2B, Molmo2-8B, LFM2.5-1.2B-Thinking, LFM2.5-1.2B-Instruct, MiniCPM5-1B, Jamba 1.7 Large, MiniCPM-V 4.6 1.3B, Kimi Linear 48B A3B Instruct, Granite 4.0 Micro, Jamba Reasoning 3B, Jamba 1.7 Mini, Phi-4 Multimodal Instruct, Tiny Aya Global, LFM2.5-VL-1.6B, Granite 4.0 H 350M, K2 Horizon 0.9B, Qwen3.5 0.8B, Granite 4.0 350M, Granite 4.0 H 1B, Molmo 7B-D, Gemma 3 270M.

80 models with no price data — shown in the strip at the left edge

Ternary Bonsai 2 27B, DeepSeek V4.1 Flash NVFP4, GLM-5.3 Flash NVFP4, GLM-5.3 NVFP4, Qwen3.8 Flash-Next NVFP4, Qwen3.8 27B NVFP4, Motif 3, Nex-N2-Pro, Ternary Bonsai 27B, K2 Horizon 375B A23B, A.X-K2, Solar Open2 250B, Bonsai 27B, Xing4.0-29B-A4B, K2-Horizon-32B, K2 Horizon MoVA 36B A4B, Maple-Preview, Intern-S2-Preview (35B-A3B), K-EXAONE 2.0 0803, G9v3-39A5B, K2 Horizon 7B, Phi-4-reasoning-plus, EXAONE 4.5 33B, Nanbeige4.1-3B, HyperNova 60B 2605, Nemotron Cascade 2 30B A3B, INTELLECT-3, Apriel-v1.6-15B-Thinker, Command A+, K2 Horizon 3.7B, Ling 3.0 Tiny, K2 Think V2, Step3 VL 10B, DiffusionGemma 26B A4B, North Mini Code, K2-V2, Falcon-H1R-7B, Solar Open 100B, Llama 3.1 Nemotron Ultra 253B v1, MiniCPM5-2B, LongCat Flash Lite, HyperCLOVA X SEED Think (32B), Tri-21B-Think, EXAONE 4.0 32B, Olmo 3.1 32B Think, LFM2.5-2.6B, Devstral 2, LFM2.5-8B-A1B, Olmo 3.1 32B Instruct, Olmo 3 7B Think, NVIDIA Nemotron 3 Nano 4B, Devstral Small 2, Phi-4-mini-flash-reasoning, Qwen3.5 2B, Falcon-H1-34B-Instruct, Hermes 4 - Llama-3.1 70B, LFM2 24B A2B, Exaone 4.0 1.2B, Llama 3.2 Instruct 90B (Vision), Gemma 4 E2B, Molmo2-8B, LFM2.5-1.2B-Thinking, LFM2.5-1.2B-Instruct, MiniCPM5-1B, Jamba 1.7 Large, MiniCPM-V 4.6 1.3B, Kimi Linear 48B A3B Instruct, Granite 4.0 Micro, Jamba Reasoning 3B, Jamba 1.7 Mini, Phi-4 Multimodal Instruct, Tiny Aya Global, LFM2.5-VL-1.6B, Granite 4.0 H 350M, K2 Horizon 0.9B, Qwen3.5 0.8B, Granite 4.0 350M, Granite 4.0 H 1B, Molmo 7B-D, Gemma 3 270M

  • Best-value frontier (nothing is both cheaper and better)
  • Evidencestrong → weak
  • Estimated from other categories
  • DeepSeek V4.1 Flash
    Quality
    75.8
    Rank
    #25/193
    Price
    $0.52/M tok
    Speed
    232tok/s
    In$0.30▾Out$1.20▾/M tok

    vs Hy4 preview: −3.1 quality · 0.42× price

  • GLM-5.3 NVFP4
    Quality
    72.7
    Rank
    #32/193
    Price
    —
    Speed
    —

    vs Hy4 preview: −6.2 quality · price n/a

  • Hy4 preview
    Quality
    78.9
    Rank
    #19/193
    Price
    $1.25/M tok
    Speed
    —
    In$0.834Out$2.501/M tok

    vs DeepSeek V4.1 Flash: +3.1 quality · 2.4× price

Overview

Metric
DeepSeek V4.1 Flash
GLM-5.3 NVFP4
Hy4 preview
Lab
DeepSeek
NVIDIA
Tencent Hunyuan
Released
Sep 10, 2026
Sep 14, 2026
Aug 28, 2026
Weights
Open
Open
Open
Input price
$/M tok
$0.30Best
—
$0.834
Output price
$/M tok
$1.20Best
—
$2.501
Blended price
$/M tok · 3:1 in:out
$0.525Best
—
$1.251
Output speed
tok/s
232
—
—
Context
tokens
1M
1M
1MBest

Quality by category

0–100 within each category (100 = best tracked model). Rank is among all models ranked there. Click a row to focus it.

Metric
DeepSeek V4.1 Flash
GLM-5.3 NVFP4
Hy4 preview
Coding
64.7±6.4
#19/114Mixedbest value
66.1±8.4
#18/114Lab-reported
66.2±8.6Best
#17/114Lab-reported
Writing & chat
69.0±4.3
#33/155Verified
72.0±9.7Best
#27/155Lab-reported
Not ranked
Vision
56.4±4.8
#42/90Verified
Not ranked
Not ranked
Hard reasoning
75.8±4.0
#25/193Mixed
72.7±7.5
#32/193Lab-reported
78.9±6.6Best
#19/193Lab-reported
Agents
82.0±7.8
#14/176Verified
Not ranked
Not ranked

Hard reasoning benchmarks

Rank among models with a published score, and the leaderboard around each model. ~ italic = no published score, estimated from related benchmarks.

Metric
DeepSeek V4.1 Flash
GLM-5.3 NVFP4
Hy4 preview
GPQA Diamond
day 0
#27/17990.9%Lab-reported
  1. #1GPT-6 Astra96.1%
  2. #26Qwen3.8-Omni-Flash91.0%
  3. #27DeepSeek V4.1 Flash90.9%
  4. #28Qwen3.8 27B90.5%
huggingface.co · 2026-09-10
#15/17992.7%BestLab-reported
  1. #1GPT-6 Astra96.1%
  2. #13Qwen3.8 Max (0902)92.8%
  3. #15GLM-5.3 NVFP492.7%
  4. #16GPT-5.6 Terra92.5%
huggingface.co · 2026-09-29
#17/17992.3%Lab-reported
  1. #1GPT-6 Astra96.1%
  2. #16GPT-5.6 Terra92.5%
  3. #17Hy4 preview92.3%
  4. #17Qwen3.8 Flash-Next92.3%
huggingface.co · 2026-08-28
Humanity's Last Exam
day 0
#26/18139.2%Verified
  1. #1Claude Opus 5.561.4%
  2. #25GLM-5.3 Flash39.9%
  3. #26DeepSeek V4.1 Flash39.2%
  4. #27MiniMax-M339.0%
artificialanalysis.ai · 2026-09-24
~#44/182~29.7%Estimated
  1. #1Claude Opus 5.561.4%
  2. #43Inkling31.9%
  3. ~#44GLM-5.3 NVFP4Estimated29.7%
  4. #44A.X-K229.6%
estimated from related benchmarks
#16/18143.4%BestLab-reported
  1. #1Claude Opus 5.561.4%
  2. #15Grok 4.644.1%
  3. #16Hy4 preview43.4%
  4. #17Grok 4.743.1%
huggingface.co · 2026-08-28
AIME (latest)
day 0
~#3/30~96.2%Estimated
  1. #1GLM-5.3 NVFP499.5%
  2. #2Inkling97.1%
  3. ~#3DeepSeek V4.1 FlashEstimated96.2%
  4. #3Ternary Bonsai 2 27B95.8%
estimated from related benchmarks
#1/2999.5%Lab-reported
  1. #1GLM-5.3 NVFP499.5%
  2. #2Inkling97.1%
huggingface.co · 2026-09-29
~#2/30~99.0%Estimated
  1. #1GLM-5.3 NVFP499.5%
  2. ~#2Hy4 previewEstimated99.0%
  3. #2Inkling97.1%
estimated from related benchmarks
AA Intelligence Index
third-party
#21/7739.5Verified
  1. #1Claude Opus 5.557.6
  2. #20Qwen3.8 Flash-Next39.8
  3. #21DeepSeek V4.1 Flash39.5
  4. #22GPT-5.538.4
artificialanalysis.ai · 2026-09-24
~#27/78~33.1Estimated
  1. #1Claude Opus 5.557.6
  2. #26Qwen3.8 27B33.7
  3. ~#27GLM-5.3 NVFP4Estimated33.1
  4. #27K2 Horizon 375B A23B30.5
estimated from related benchmarks
~#24/78~37.5Estimated
  1. #1Claude Opus 5.557.6
  2. #23Claude Sonnet 538.2
  3. ~#24Hy4 previewEstimated37.5
  4. #24GPT-6 Luna37.3
estimated from related benchmarks
Benchmarks in other categories (4)

Coding benchmarks

Rank among models with a published score, and the leaderboard around each model. ~ italic = no published score, estimated from related benchmarks.

Metric
DeepSeek V4.1 Flash
GLM-5.3 NVFP4
Hy4 preview
SWE-bench Verified
day 0
~#3/14~83.2%Estimated
  1. #1Claude Opus 596.0%
  2. #2Claude Sonnet 585.2%
  3. ~#3DeepSeek V4.1 FlashEstimated83.2%
  4. #3Gemini 3.1 Pro80.6%
estimated from related benchmarks
~#2/14~88.0%Estimated
  1. #1Claude Opus 596.0%
  2. ~#2GLM-5.3 NVFP4Estimated88.0%
  3. #2Claude Sonnet 585.2%
estimated from related benchmarks
~#3/14~85.1%Estimated
  1. #1Claude Opus 596.0%
  2. #2Claude Sonnet 585.2%
  3. ~#3Hy4 previewEstimated85.1%
  4. #3Gemini 3.1 Pro80.6%
estimated from related benchmarks
SWE-bench Pro (Public)
day 0
~#6/19~65.8%Estimated
  1. #1Claude Opus 5.589.9%
  2. #5Intern-S2-397B68.5%
  3. ~#6DeepSeek V4.1 FlashEstimated65.8%
  4. #6Hy4 preview65.7%
estimated from related benchmarks
~#5/19~71.4%Estimated
  1. #1Claude Opus 5.589.9%
  2. #4Claude Opus 579.2%
  3. ~#5GLM-5.3 NVFP4Estimated71.4%
  4. #5Intern-S2-397B68.5%
estimated from related benchmarks
#6/1865.7%Lab-reported
  1. #1Claude Opus 5.589.9%
  2. #5Intern-S2-397B68.5%
  3. #6Hy4 preview65.7%
  4. #7GPT-5.6 Terra63.4%
huggingface.co · 2026-08-28
DeepSWE v1.1
day 0
#2/3174.2%BestLab-reported
  1. #1GPT-6.1 Sol75.2%
  2. #2DeepSeek V4.1 Flash74.2%
  3. #3Grok 4.772.6%
huggingface.co · 2026-09-10
~#20/32~63.4%Estimated
  1. #1GPT-6.1 Sol75.2%
  2. #19GPT-6 Luna63.7%
  3. ~#20GLM-5.3 NVFP4Estimated63.4%
  4. #20GLM-5.3 Flash63.4%
estimated from related benchmarks
#17/3164.3%Lab-reported
  1. #1GPT-6.1 Sol75.2%
  2. #17Claude Fable 5.164.3%
  3. #17Hy4 preview64.3%
  4. #19GPT-6 Luna63.7%
huggingface.co · 2026-08-28
Terminal-Bench 4.0
day 0
#15/8426.8%Verified
  1. #1Claude Sonnet 5.563.6%
  2. #14GLM-5.3 Flash32.8%
  3. #15DeepSeek V4.1 Flash26.8%
  4. #16Grok 4.725.8%
artificialanalysis.ai · 2026-09-24
~#15/85~27.3%Estimated
  1. #1Claude Sonnet 5.563.6%
  2. #14GLM-5.3 Flash32.8%
  3. ~#15GLM-5.3 NVFP4Estimated27.3%
  4. #15DeepSeek V4.1 Flash26.8%
estimated from related benchmarks
~#15/85~27.2%Estimated
  1. #1Claude Sonnet 5.563.6%
  2. #14GLM-5.3 Flash32.8%
  3. ~#15Hy4 previewEstimated27.2%
  4. #15DeepSeek V4.1 Flash26.8%
estimated from related benchmarks
LiveCodeBench
third-party
~#9/25~79.4%Estimated
  1. #1Qwen3.8-Omni-Flash92.6%
  2. #8Gemma 4 31B80.0%
  3. ~#9DeepSeek V4.1 FlashEstimated79.4%
  4. #9gpt-oss-20b77.7%
estimated from related benchmarks
#14/2453.5%Lab-reported
  1. #1Qwen3.8-Omni-Flash92.6%
  2. #13Claude Haiku 4.561.5%
  3. #14GLM-5.3 NVFP453.5%
  4. #15Phi-4-reasoning-plus53.1%
huggingface.co · 2026-09-29
~#7/25~82.9%Estimated
  1. #1Qwen3.8-Omni-Flash92.6%
  2. #6gpt-oss-120b87.8%
  3. ~#7Hy4 previewEstimated82.9%
  4. #7Ternary Bonsai 27B82.8%
estimated from related benchmarks
SciCode
third-party
#28/10251.9%Verified
  1. #1Claude Opus 5.566.9%
  2. #26Qwen3.8 Max (0902)52.1%
  3. #28DeepSeek V4.1 Flash51.9%
  4. #29GLM-5.3 Flash51.6%
artificialanalysis.ai · 2026-09-24
#10/10258.0%BestLab-reported
  1. #1Claude Opus 5.566.9%
  2. #9Gemini 3.1 Pro58.7%
  3. #10GLM-5.3 NVFP458.0%
  4. #11GLM-5.3 Flash NVFP457.7%
huggingface.co · 2026-09-29
~#21/103~54.7%Estimated
  1. #1Claude Opus 5.566.9%
  2. #20GPT-5.6 Terra55.0%
  3. ~#21Hy4 previewEstimated54.7%
  4. #21GPT-6 Luna54.6%
estimated from related benchmarks
AA Coding Agent Index
third-party
~#10/15~48.0Estimated
  1. #1Claude Opus 5.566.0
  2. #9Kimi K351.9
  3. ~#10DeepSeek V4.1 FlashEstimated48.0
  4. #10Grok 4.647.0
estimated from related benchmarks
~#6/15~56.5Estimated
  1. #1Claude Opus 5.566.0
  2. #5GPT-6 Sol56.7
  3. ~#6GLM-5.3 NVFP4Estimated56.5
  4. #6Grok 4.756.3
estimated from related benchmarks
~#9/15~53.0Estimated
  1. #1Claude Opus 5.566.0
  2. #8GLM-5.353.6
  3. ~#9Hy4 previewEstimated53.0
  4. #9Kimi K351.9
estimated from related benchmarks

Writing & chat benchmarks

Rank among models with a published score, and the leaderboard around each model. ~ italic = no published score, estimated from related benchmarks.

Metric
DeepSeek V4.1 Flash
GLM-5.3 NVFP4
Hy4 preview
IFBench
day 0
~#27/134~72.2%Estimated
  1. #1MiniMax-M382.9%
  2. #26Gemma 4 26B A4B72.4%
  3. ~#27DeepSeek V4.1 FlashEstimated72.2%
  4. #27Nemotron 3 Super 120B A12B71.5%
estimated from related benchmarks
#43/13366.0%Lab-reported
  1. #1MiniMax-M382.9%
  2. #41Nova 2.0 Omni66.2%
  3. #43GLM-5.3 NVFP466.0%
  4. #43Olmo 3.1 32B Think66.0%
huggingface.co · 2026-09-29
—
MMMLU
day 0
~#2/5~90.8%Estimated
  1. #1Gemini 3.1 Pro92.6%
  2. ~#2DeepSeek V4.1 FlashEstimated90.8%
  3. #2Gemma 4 31B88.4%
estimated from related benchmarks
~#3/5~87.7%Estimated
  1. #1Gemini 3.1 Pro92.6%
  2. #2Gemma 4 31B88.4%
  3. ~#3GLM-5.3 NVFP4Estimated87.7%
  4. #3Claude Haiku 4.583.0%
estimated from related benchmarks
—
LMArena Text, Non-English (Elo)
third-party
#12/531462Verified
  1. #1Claude Opus 5.51501
  2. #11GLM-5.3 Flash1463
  3. #12DeepSeek V4.1 Flash1462
  4. #13MiMo-V2.6-Pro1461
lmarena.ai · 2026-09-25
~#20/54~1446Estimated
  1. #1Claude Opus 5.51501
  2. #18Grok 4.61447
  3. ~#20GLM-5.3 NVFP4Estimated1446
  4. #20Gemini 3.5 Flash-Lite1445
estimated from related benchmarks
—
LMArena Text (Elo)
third-party
#13/531477Verified
  1. #1Claude Opus 5.51509
  2. #10MiMo-V2.6-Pro1480
  3. #13DeepSeek V4.1 Flash1477
  4. #14GLM-5.3 Flash1475
lmarena.ai · 2026-09-25
~#18/54~1458Estimated
  1. #1Claude Opus 5.51509
  2. #17Claude Sonnet 51461
  3. ~#18GLM-5.3 NVFP4Estimated1458
  4. #18GPT-6 Sol1457
estimated from related benchmarks
—
EQ-Bench Creative Writing v3 (Elo)
third-party
#20/251540Verified
  1. #1GPT-6 Astra2173
  2. #19Gemini 3.5 Flash-Lite1559
  3. #20DeepSeek V4.1 Flash1540
  4. #21Gemini 3.1 Pro1491
eqbench.com · 2026-09-24
~#9/26~1988Estimated
  1. #1GPT-6 Astra2173
  2. #8Grok 4.72007
  3. ~#9GLM-5.3 NVFP4Estimated1988
  4. #9Muse Spark 1.31906
estimated from related benchmarks
—

Vision benchmarks

Rank among models with a published score, and the leaderboard around each model. ~ italic = no published score, estimated from related benchmarks.

Metric
DeepSeek V4.1 Flash
GLM-5.3 NVFP4
Hy4 preview
MMMU
day 0
~#2/3~71.3%Estimated
  1. #1Claude Haiku 4.573.2%
  2. ~#2DeepSeek V4.1 FlashEstimated71.3%
  3. #2Llama 4 Maverick69.4%
estimated from related benchmarks
—
—
MMMU-Pro
day 0
#24/8477.0%Verified
  1. #1Claude Opus 5.587.7%
  2. #22Qwen3.5 397B A17B77.3%
  3. #24DeepSeek V4.1 Flash77.0%
  4. #25Intern-S2-Preview (35B-A3B)76.9%
artificialanalysis.ai · 2026-09-24
—
—
CharXiv Reasoning
day 0
~#5/9~79.2%Estimated
  1. #1Gemini 3.8 Flash86.2%
  2. #4Qwen3.8-Omni-Flash83.5%
  3. ~#5DeepSeek V4.1 FlashEstimated79.2%
  4. #5Muse Glimmer78.8%
estimated from related benchmarks
—
—
OmniDocBench v1.5
third-party
~#5/8~85.2%Estimated
  1. #1MiniMax-M391.6%
  2. #4Gemini 3.1 Pro85.3%
  3. ~#5DeepSeek V4.1 FlashEstimated85.2%
  4. #5Claude Haiku 4.579.6%
estimated from related benchmarks
—
—
GDP.pdf (AA)
third-party
#24/4212.8%Verified
  1. #1GPT-6 Astra31.0%
  2. #23Claude Sonnet 513.2%
  3. #24DeepSeek V4.1 Flash12.8%
  4. #24Inkling12.8%
artificialanalysis.ai · 2026-09-24
—
—
Chartography
third-party
#19/2211.6%Verified
  1. #1GPT-6 Astra71.0%
  2. #18Grok 4.714.7%
  3. #19DeepSeek V4.1 Flash11.6%
  4. #20Gemini 3.5 Flash-Lite10.2%
surgehq.ai · 2026-09-24
—
—

Agents benchmarks

Rank among models with a published score, and the leaderboard around each model. ~ italic = no published score, estimated from related benchmarks.

Metric
DeepSeek V4.1 Flash
GLM-5.3 NVFP4
Hy4 preview
τ²-bench (Telecom)
day 0
~#5/117~94.7%Estimated
  1. #1Step 3.7 Flash98.5%
  2. #4Qwen3.6 35B A3B95.3%
  3. ~#5DeepSeek V4.1 FlashEstimated94.7%
  4. #5Mistral Medium 3.594.2%
estimated from related benchmarks
—
—
OSWorld-Verified
day 0
~#5/12~81.7%Estimated
  1. #1Kimi K384.8%
  2. #4MiMo-V2.6-Pro82.0%
  3. ~#5DeepSeek V4.1 FlashEstimated81.7%
  4. #5Claude Sonnet 581.2%
estimated from related benchmarks
—
—
OSWorld 2.0
day 0
~#3/5~57.7%Estimated
  1. #1GLM-5.3 Flash59.1%
  2. #2Kimi K358.3%
  3. ~#3DeepSeek V4.1 FlashEstimated57.7%
  4. #3GPT-5.6 Terra50.2%
estimated from related benchmarks
—
—
BrowseComp
day 0
~#7/18~87.6%Estimated
  1. #1Atria Dawn Preview92.5%
  2. #6Step 5 Preview88.7%
  3. ~#7DeepSeek V4.1 FlashEstimated87.6%
  4. #7GPT-5.6 Terra87.5%
estimated from related benchmarks
—
—
GDPval-AA (Elo)
third-party
#12/861600Verified
  1. #1Claude Opus 5.51846
  2. #11Qwen3.8 Flash-Next1612
  3. #12DeepSeek V4.1 Flash1600
  4. #13Qwen3.8 2.4T A95B1597
artificialanalysis.ai · 2026-09-24
—
—
AutomationBench-AA
third-party
#2/7968.9%Verified
  1. #1Claude Opus 5.569.5%
  2. #2DeepSeek V4.1 Flash68.9%
  3. #3GPT-6 Astra68.5%
artificialanalysis.ai · 2026-09-24
—
—
τ-Bench Banking (AA)
third-party
~#14/90~40.6%Estimated
  1. #1Muse Spark 1.350.5%
  2. #13GPT-6 Astra41.4%
  3. ~#14DeepSeek V4.1 FlashEstimated40.6%
  4. #14GPT-5.6 Terra40.2%
estimated from related benchmarks
—
—