Skip to content
aiqbench

DeepSeek · model passport

DeepSeek: DeepSeek V3.1 Terminus

deepseek-v3-1-terminus
  • Open weights
  • MIT licence
  • 684.5B params
  • 163,840 context
  • Released 2025-09-22(curated)
8/27axes covered
Overall · LMArena
1,415#129 of 396
Best percentile
92.1MMLU-Pro (Artificial Analysis)
Cohort standing
66.1median 52.6

What the data says

Benchmarked on 8 of 27 tracked axes. Every figure below is the model's own published result on that board.

Strongest axes

  • 92.1 pct — MMLU-Pro (Artificial Analysis) (85.1%)
  • 90.1 pct — LiveCodeBench (Artificial Analysis) (79.8%)

Uneven: 31.2 percentile points between its best and worst axis — 92.1 pct on MMLU-Pro (Artificial Analysis) versus 60.9 pct on Coding.

  • MMLU-Pro (Artificial Analysis)92.1 pct85.1%Verify →
  • LiveCodeBench (Artificial Analysis)90.1 pct79.8%Verify →
  • LiveBench72.2 pct64.7%Verify →
  • Overall67.7 pct1,415Verify →
  • IFBench (Artificial Analysis)65.7 pct57.0%Verify →
  • Hard prompts64.4 pct1,424Verify →
  • Instruction following63.9 pct1,394Verify →
  • Coding60.9 pct1,440Verify →
  • WebDev ArenaNot benchmarked
  • Vision ArenaNot benchmarked
  • Factuality ArenaNot benchmarked
  • Search ArenaNot benchmarked
  • Document ArenaNot benchmarked
  • Agent ArenaNot benchmarked
  • GPQA DiamondNot benchmarked
  • AIME (math)Not benchmarked
  • FrontierMathNot benchmarked
  • SimpleQA VerifiedNot benchmarked
  • ARC-AGI-2Not benchmarked
  • Humanity's Last ExamNot benchmarked
  • SimpleBenchNot benchmarked
  • BFCL v4 (function calling)Not benchmarked
  • SWE-bench Verified (live board)Not benchmarked
  • Terminal-Bench 4.0Not benchmarked
  • HHEM Hallucination (Vectara)Not benchmarked
  • SWE-bench VerifiedNot benchmarked
  • Terminal-BenchNot benchmarked

Percentiles are the model's standing within each board's own population. Axes marked "not benchmarked" have no published result for this model — that is an absence of data, not a low score.

How to run it

Run it yourself

Repositorydeepseek-ai/DeepSeek-V3.1-Terminus
Parameters684.5B
ArchitectureDeepseekV3ForCausalLM
LicenceMIT

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "deepseek-ai/DeepSeek-V3.1-Terminus"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

Ollama

We don't sync Ollama tags, so we won't print a command we can't verify — a wrong one would pull the wrong model. Search the Ollama library →

Unsloth fine-tuning recipes are model-specific and we don't derive them here. Unsloth docs →

Use it hosted

API model IDdeepseek/deepseek-v3.1-terminus
Context length163,840 tokens
Max output147,456 tokens
Input price$0.27 per 1M tokens
Output price$1 per 1M tokens

API parameters this endpoint supports

Frequency penaltyInclude reasoningMax tokensReasoningResponse formatStop sequencesStructured outputsTemperatureTool choiceTool callingTop-kTop-p

Example request

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek/deepseek-v3.1-terminus","messages":[{"role":"user","content":"Hello"}]}'

View on OpenRouter →

Identifiers, limits and prices are synced from the upstream source shown above. Any field we haven't synced is left out entirely rather than guessed.

  • deepseek-v3.1-terminus stands at 66.1 (raw mean 72.1) — measured on 8 of 27 axes.

    standingtemplateVerify →

  • deepseek-v3.1-terminus stands at 66.1 versus a cohort median of 52.6 across 143 models released within 180 days — above the peer median.

    cohorttemplateVerify →

Every claim is a computed fact from the current board data, phrased by a low-temperature model call and rejected by a validation gate if any number is not in the source row. How the leaderboards work →