ResearchLLM Verifications Leaderboard

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

Head to head · tested 27 Oct 2026

Llama 3.3 70B Instruct vs Ministral 3 8B 2512: accuracy, speed and price

Similar accuracy; Ministral 3 8B 2512 costs less per question.

MetaLlama 3.3 70B Instruct42.5% score · #52$0.00053 per question42 tok/sMistralMinistral 3 8B 251242.5% score · #52$0.00027 per question124 tok/s

Verdict

Built from the measurements below

  • Accuracy is too close to call: Llama 3.3 70B Instruct scored 42.5% and Ministral 3 8B 2512 42.5% on the same 80 questions, a gap within what chance alone could produce.
  • Ministral 3 8B 2512 is 2.0× cheaper per question as billed ($0.00027 vs $0.00053), reasoning tokens included.
  • Ministral 3 8B 2512 generates text 3.0× faster: 124 tokens per second against 42.
  • Ministral 3 8B 2512 starts answering sooner: 0.3 s to the first token, against 0.7 s for Llama 3.3 70B Instruct.
  • Ministral 3 8B 2512 accepts more context: 262k tokens against 131k.

Which should you choose?

Choose Llama 3.3 70B Instruct for shorter outputs (about 855 tokens per question against 1,496) and more hosts to choose from on OpenRouter (9 against 1). Choose Ministral 3 8B 2512 for the lower cost per correct answer ($0.00063 against $0.0013), faster streaming, a quicker first token and a 262k-token context window. The widest subject gap is in economics: Ministral 3 8B 2512 answered 7 of 9 correctly, Llama 3.3 70B Instruct 5 of 9. With so few questions per subject, treat this as a lead to test on your own prompts, not a result.

Accuracy gap +0.0 points (z = 0.00). A winner is called only when z is at least 1.96, the usual 95% threshold. Speed, latency and price need a 10% margin.

Head to head

Same questions, same prompt, same API

MeasureLlama 3.3 70B InstructMinistral 3 8B 2512Edge
Verification scoreHigher is better42.5%42.5%Too close to call
Cost per questionLower is better$0.00053$0.00027Ministral 3 8B 2512
Output speedHigher is better42124Ministral 3 8B 2512
Time to first tokenLower is better0.7 s0.3 sMinistral 3 8B 2512
End-to-end response timeLower is better8.8 s3.5 sMinistral 3 8B 2512
Tokens per questionLower is better8551,496Llama 3.3 70B Instruct
Context windowHigher is better131k262kMinistral 3 8B 2512

Side by side

Bars are scaled to the larger of the two values. The winner on each measure is marked in bold.

Every measure

Llama 3.3 70B Instruct first, Ministral 3 8B 2512 second

Verification scorehigher is better
Llama 3.3 70B Instruct42.5%
Ministral 3 8B 251242.5%
Cost per questionlower is better
Llama 3.3 70B Instruct$0.00053
Ministral 3 8B 2512$0.00027
Output speedhigher is better
Llama 3.3 70B Instruct42
Ministral 3 8B 2512124
Time to first tokenlower is better
Llama 3.3 70B Instruct0.7 s
Ministral 3 8B 25120.3 s
End-to-end response timelower is better
Llama 3.3 70B Instruct8.8 s
Ministral 3 8B 25123.5 s
Tokens per questionlower is better
Llama 3.3 70B Instruct855
Ministral 3 8B 25121,496
Context windowhigher is better
Llama 3.3 70B Instruct131k
Ministral 3 8B 2512262k

Score vs. cost per question

Both 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 4Ministral 3 8B 2512Granite 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

Specifications and pricing

From OpenRouter's catalogue, plus the leaderboard's test dates

SpecificationLlama 3.3 70B InstructMinistral 3 8B 2512
CreatorMetaMistral
Released6 Dec 20242 Dec 2025
WeightsOpen weightsOpen weights
ReasoningNoNo
Inputtexttext, image
Outputtexttext
Context window131k tokens262k tokens
Knowledge cutoff31 Dec 2023—
List price in / out per 1M$0.1 / $0.32$0.15 / $0.15
Blended price per 1M (3:1)$0.16$0.15
Cheapest hostDeepInfra · $0.1 / $0.32Mistral · $0.15 / $0.15
Hosts on OpenRouter91
Last tested26 Oct 202626 Oct 2026

Compare any two models →

Llama 3.3 70B Instruct vs Ministral 3 8B 2512: frequently asked questions

Is Llama 3.3 70B Instruct better than Ministral 3 8B 2512?
On accuracy they are statistically tied: Llama 3.3 70B Instruct scored 42.5% and Ministral 3 8B 2512 42.5% on the same 80 questions from open datasets. Similar accuracy; Ministral 3 8B 2512 costs less per question.
Which is cheaper, Llama 3.3 70B Instruct or Ministral 3 8B 2512?
Ministral 3 8B 2512 is cheaper: $0.00027 per question as billed, against $0.00053. List prices: Llama 3.3 70B Instruct $0.1 in / $0.32 out, Ministral 3 8B 2512 $0.15 in / $0.15 out per million tokens.
Which is faster, Llama 3.3 70B Instruct or Ministral 3 8B 2512?
Ministral 3 8B 2512 is faster at 124 output tokens per second, against 42. Time to first token: 0.7 s for Llama 3.3 70B Instruct, 0.3 s for Ministral 3 8B 2512.
Which has the larger context window, Llama 3.3 70B Instruct or Ministral 3 8B 2512?
Ministral 3 8B 2512, with 262k tokens against 131k.