NOUMENOpen app
NOUMEN

DeepSeek · Open weights

DeepSeek V3.1 Terminus (Thinking)

Added Jan 1, 2024Open weights

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Thinking-enabled DeepSeek-V3.1-Terminus with improved language consistency, upgraded Code/Search Agents, and stronger stability and reliability versus V3.1. FP8.

  • Tools
  • Structured output
Input
Text
Output
Text
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.25
Output
$0.70

Reference prices as published by the model provider. API example

Specifications

Context window
128K
Max output
65.5K
Average output (7 days)
362 tokens
Longer than 37% of 468 models

Side by side

DeepSeek V3.1 Terminus (Thinking) compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

DeepSeek V3.1 Terminus (Thinking)DeepSeekThis modelDeepSeek V4.1 Flash 1017 ThinkingDeepSeekRemove
ProviderDeepSeekDeepSeek
Discovery tierOpen weightsOpen weights
AddedJan 1, 2024Oct 17, 2026
Limits
Context window128K1.05MBest
Max output65.54K393.22KBest
Average output362 tokens1.2K tokens
Pricing, USD per 1M tokens
Input$0.25$0.10Best
Output$0.70$0.60Best
Cache read—$0.003
Performance
Throughput21.6 tok/s214.9 tok/sBest
Time to first token6.4 s2.4 sBest
Request success (24h)99.5%99.8%Best
Benchmark indexes
Intelligence Index14.845.1Best
Coding Index43.5—
Math Index89.7—
Capabilities
ReasoningNoYes
ToolsYesYes
VisionNoYes
Structured outputYesYes
Reasoning effort—none, low, high, max

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • DeepSeek V3.1 Terminus (Thinking)
  • DeepSeek V4.1 Flash 1017 Thinking

Intelligence Index

  • DeepSeek V3.1 Terminus (Thinking)14.8
  • DeepSeek V4.1 Flash 1017 Thinking45.1

DeepSeek V3.1 Terminus (Thinking): Better than 37% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • DeepSeek V3.1 Terminus (Thinking): 43.5
    • DeepSeek V4.1 Flash 1017 Thinking: —

  • Math Index

    Composite of the math evaluations

    • DeepSeek V3.1 Terminus (Thinking): 89.7
    • DeepSeek V4.1 Flash 1017 Thinking: —

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • DeepSeek V3.1 Terminus (Thinking): 2.7%
    • DeepSeek V4.1 Flash 1017 Thinking: 66.6%Top

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • DeepSeek V3.1 Terminus (Thinking): 0.2%
    • DeepSeek V4.1 Flash 1017 Thinking: 36.7%Top

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • DeepSeek V3.1 Terminus (Thinking): 458 Elo
    • DeepSeek V4.1 Flash 1017 Thinking: 1,570 EloTop

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • DeepSeek V3.1 Terminus (Thinking): 709 Elo
    • DeepSeek V4.1 Flash 1017 Thinking: 1,746 EloTop

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • DeepSeek V3.1 Terminus (Thinking): 5.4%
    • DeepSeek V4.1 Flash 1017 Thinking: 17.2%Top

  • AA-LCRv1.1

    Reasoning across long documents

    • DeepSeek V3.1 Terminus (Thinking): 69.3%
    • DeepSeek V4.1 Flash 1017 Thinking: 82.7%Top

  • MLCR-AA

    Long-context reasoning over medical records

    • DeepSeek V3.1 Terminus (Thinking): —
    • DeepSeek V4.1 Flash 1017 Thinking: 26.7%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • DeepSeek V3.1 Terminus (Thinking): 16.4%
    • DeepSeek V4.1 Flash 1017 Thinking: 43.3%Top

  • GPQA Diamond

    Graduate-level science questions

    • DeepSeek V3.1 Terminus (Thinking): 79.2%
    • DeepSeek V4.1 Flash 1017 Thinking: —

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • DeepSeek V3.1 Terminus (Thinking): 85.1%
    • DeepSeek V4.1 Flash 1017 Thinking: —

  • IFBench

    Following precise output instructions

    • DeepSeek V3.1 Terminus (Thinking): 57.0%
    • DeepSeek V4.1 Flash 1017 Thinking: —

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • DeepSeek V3.1 Terminus (Thinking): 0.0%
    • DeepSeek V4.1 Flash 1017 Thinking: 33.8%Top

  • SciCode

    Scientific computing in Python

    • DeepSeek V3.1 Terminus (Thinking): 38.0%
    • DeepSeek V4.1 Flash 1017 Thinking: 55.9%Top

  • LiveCodeBench

    Recent competitive programming problems

    • DeepSeek V3.1 Terminus (Thinking): 79.8%
    • DeepSeek V4.1 Flash 1017 Thinking: —

Math

  • AIME 2025

    Competition mathematics, 2025 problems

    • DeepSeek V3.1 Terminus (Thinking): 89.7%
    • DeepSeek V4.1 Flash 1017 Thinking: —

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • DeepSeek V3.1 Terminus (Thinking): 30.3%
    • DeepSeek V4.1 Flash 1017 Thinking: —

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • DeepSeek V3.1 Terminus (Thinking): 37.1%
    • DeepSeek V4.1 Flash 1017 Thinking: —

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • DeepSeek V3.1 Terminus (Thinking): 69.3%
    • DeepSeek V4.1 Flash 1017 Thinking: 82.7%Top

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Throughput
21.6 tok/s
Faster than 14% of 254 models
Time to first token
6.4 s
Quicker than 27% of 254 models
Prompt cache hits
73.2%
Request success (24h)
99.5%
Reliable

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Hourly request success · last 72 hoursLowest hour 91.6%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-ai/DeepSeek-V3.1-Terminus:thinking",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-V3.1-Terminus:thinking",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

DeepSeek

Other DeepSeek models, starting with the same family.