ResearchLLM Verifications Leaderboard

Data as of · 78 models, 80 questions · reviewed by Suhas Bhairav · methodology

Provider

Meta AI models

5 models tested, compared with 73 models from other labs.

Best verification score78.8Muse Wonder · #7 of 78Best output speed220Muse Spark Open · #7 of 78Best time to first token0.5 sMuse Spark Open · #14 of 78Best cost per question$0.00037Llama 4 Scout · #13 of 78
Meta#7Muse Wonder78.8% verification score175 tok/s16.1 s first token$0.021 / questionMeta#9Muse Spark Open77.5% verification score220 tok/s0.5 s first token$0.0045 / questionMeta#40Llama 4 Maverick59.0% verification score13 tok/s0.7 s first token$0.00059 / questionMeta#52Llama 3.3 70B Instruct42.5% verification score42 tok/s0.7 s first token$0.00053 / questionMeta#54Llama 4 Scout41.2% verification score15 tok/s1.2 s first token$0.00037 / question

Score vs. cost per question

Meta models in color, every other model in grey

Most attractive quadrant0255075100$0.00001$0.0001$0.001$0.01$0.10Cost per question (USD, log) →Verification score (%) →Qwen3.8 2.4T A95BCommand AReka Flash 3Gemma 4 26B A4BGemma 4 31BQwen3.5 397B A17BHermes 4 405BMuse WonderGrok 4.7o3gpt-oss-20bMinistral 3 3B 2512Nova Micro 1.0Phi 4Granite 4.2 8BHy3GLM 5.3 FlashSolar Pro 4Trinity Large ThinkingMiniMax M3Mistral Large 4Claude Opus 5.5GPT-6.1 SolGemini 4 ArgonGPT-5.4GPT-6 Lunagpt-oss-120bLlama 4 Maverick

Output speed vs. cost per question

Meta models in color, every other model in grey

Most attractive quadrant0200400600800$0.00001$0.0001$0.001$0.01$0.10Cost per question (USD, log) →Output speed (tok/s) →Nemotron 3.5 Super VLQwen3.5 397B A17BMuse WonderGrok 4.7o3gpt-oss-20bMinistral 3 3B 2512Nova Micro 1.0Trinity Large ThinkingInkling SmallGemini 3.5 Flash LiteMuse Spark OpenGPT-6 Luna

Verification score

Meta models highlighted against every model in this run · higher is better

  1. 1Claude Opus 5.585.0
  2. 2Gemini 4 Argon83.8
  3. 3GPT-5.3-Codex82.5
  4. 3Gemini 3.8 Flash82.5
  5. 5Claude Fable 5.181.2
  6. 6Grok 4.780.0
  7. 7Muse Wonder78.8
  8. 7Claude Sonnet 5.578.8
  9. 9Gemini 3.1 Pro Preview77.5
  10. 9Muse Spark Open77.5
  11. 40Llama 4 Maverick59.0*
  12. 52Llama 3.3 70B Instruct42.5
  13. 54Llama 4 Scout41.2

* Llama 4 Maverick: Scored on 78 of 80 questions: 2 failed on the host's side and are left out.

59 more models · see the full leaderboard

6 of 78 tested models have no verification score

Output speed

Meta models highlighted against every model in this run · higher is better

  1. 1Nemotron 3.5 Super VL715
  2. 2Gemini 3.5 Flash Lite409
  3. 3Trinity Large Thinking329
  4. 4Ministral 3 3B 2512252
  5. 5Qwen3.5 397B A17B240
  6. 6Inkling Small233
  7. 7Muse Spark Open220
  8. 8Claude Haiku 5.5192
  9. 9Nova Micro 1.0180
  10. 10Muse Wonder175
  11. 63Llama 3.3 70B Instruct42
  12. 69Llama 4 Scout15
  13. 70Llama 4 Maverick13

59 more models · see the full leaderboard

6 of 78 tested models have no output speed

Time to first token

Meta models highlighted against every model in this run · lower is better

  1. 1Ministral 3 3B 25120.3 s
  2. 2Ministral 3 8B 25120.3 s
  3. 3Hermes 4 405B0.3 s
  4. 4Kimi K2.7 Code0.4 s
  5. 5MiMo-V2.6-Flash0.4 s
  6. 6Qwen3.6 35B A3B0.4 s
  7. 7Qwen3 Coder Next0.4 s
  8. 8Mistral Medium 3.50.5 s
  9. 9Phi 40.5 s
  10. 10Nova Micro 1.00.5 s
  11. 14Muse Spark Open0.5 s
  12. 30Llama 4 Maverick0.7 s
  13. 33Llama 3.3 70B Instruct0.7 s
  14. 50Llama 4 Scout1.2 s
  15. 73Muse Wonder16.1 s

59 more models · see the full leaderboard

4 of 78 tested models have no time to first token

Cost per question

Meta models highlighted against every model in this run · lower is better

  1. 1Granite 4.0 Micro$0.000051
  2. 2gpt-oss-20b$0.000066
  3. 3Granite 4.2 8B$0.000077
  4. 4Nova Micro 1.0$0.000087
  5. 5Mercury 2.5$0.00011
  6. 6Phi 4$0.00012
  7. 7gpt-oss-120b$0.00013
  8. 8GLM 5.3 Flash$0.00015
  9. 9Ministral 3 3B 2512$0.00019
  10. 10GPT-6 Luna$0.00025
  11. 13Llama 4 Scout$0.00037
  12. 18Llama 3.3 70B Instruct$0.00053
  13. 20Llama 4 Maverick$0.00059
  14. 48Muse Spark Open$0.0045
  15. 70Muse Wonder$0.021

57 more models · see the full leaderboard

6 of 78 tested models have no cost per question
  • LFM2.5-2.6B: no verification score (80 of 80 questions failed on the host's side)
  • Ling 3.0 Flash VL: no verification score (not yet run on the current question suite)
  • Qwen3.5-122B-A10B: no verification score (not yet run on the current question suite)
  • Qwen3 Next 80B A3B Thinking: no verification score (not yet run on the current question suite)
  • Nova Premier 1.0: no verification score (not yet run on the current question suite)
  • GPT-5.5 Pro: no verification score (not yet run on the current question suite)

End-to-end response time

Meta models highlighted against every model in this run · lower is better

  1. 1Nemotron 3.5 Super VL1.5 s
  2. 2Ministral 3 3B 25121.6 s
  3. 3Nova Micro 1.02.1 s
  4. 4Trinity Large Thinking2.7 s
  5. 5Gemini 3.5 Flash Lite2.8 s
  6. 6DeepSeek V4 Flash Vision Exp3.2 s
  7. 7Muse Spark Open3.2 s
  8. 8GLM 5.33.4 s
  9. 9Ministral 3 8B 25123.5 s
  10. 10Qwen3.5 397B A17B3.5 s
  11. 48Llama 3.3 70B Instruct8.8 s
  12. 69Muse Wonder20.8 s
  13. 70Llama 4 Scout25.6 s
  14. 71Llama 4 Maverick27.9 s

60 more models · see the full leaderboard

4 of 78 tested models have no end-to-end response time

All Meta results

Leaderboard
#
1Muse Wonder
Meta · served by Meta
78.863 / 8017516.1 s$0.021$4 / $201M
2Muse Spark Open
Meta · served by Together
77.562 / 802200.5 s$0.0045$1.25 / $4.251M
3Llama 4 Maverick
Meta · served by DigitalOcean
59.0*46 / 78130.7 s$0.00059$0.1875 / $0.6525524k
4Llama 3.3 70B Instruct
Meta · served by DeepInfra
42.534 / 80420.7 s$0.00053$0.1 / $0.32131k
5Llama 4 Scout
Meta · served by DeepInfra
41.233 / 80151.2 s$0.00037$0.1 / $0.31.3M

Meta: quick answers

What is the best Meta model?
Muse Wonder has the highest verification score of the Meta models the leaderboard tested: 78.8% (63 of 80 correct) in the 2026-10-27 run.
What is the fastest Meta model?
Muse Spark Open is the fastest Meta model the leaderboard tested, at a median 220 output tokens per second.
What is the cheapest Meta model?
Llama 4 Scout has the lowest billed cost per question of the Meta models the leaderboard tested: $0.00037.
Which Meta models do you test?
Muse Wonder, Muse Spark Open, Llama 4 Maverick, Llama 3.3 70B Instruct, Llama 4 Scout.