Skip to content

DeepSeek · Chat

DeepSeek V3.1 Terminus

DeepSeek 🇨🇳ChatDeepSeek familyOne of its sources says the weights aren't publicly available.
From NanoGPT's listing:
DeepSeek-V3.1-Terminus. The latest update builds on V3.1's strengths while addressing key user feedback. Language consistency improvements (fewer CN/EN mix-ups, no random chars), stronger Code Agent & Search Agent performance, and more stable, reliable outputs across benchmarks. FP8.
llminfo IQ
121.4
Maker's price
$0.44 blended / 1M
Max context
164K
Listings
23

Evidence and sources

🤗 deepseek-ai/DeepSeek-V3.1-Terminusllminfo IQ 121.4, ranked among chat models. From 57 benchmarks. Plenty of data behind it.How llminfo IQ is computedCoding IQ 99.4, 95% interval 86–112. From 3 benchmark families across 2 sources. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.How Coding IQ is computedGood value (57.6). Cheaper per task than its benchmark results predict. From 5 benchmarks with measured task costs. A fair amount of data behind it.How Value for money worksYap score: tokens used per coding task vs. models with the same resultsTypical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.Median 18,032 output tokens per task at default effort.How the Yap score worksReleased 22 Sep 2025 Training-data cutoff, as its sources state it Maker HQ: CN (according to DeepSeek's own website)
All rankings: #136 on NanoGPT recently #251 on OpenRouter today #263 on OpenRouter this week #269 on OpenRouter this monthSources: LMArena data, SiliconFlow data, Models.dev data, OpenRouter data docs, NanoGPT, Hugging FacePrimary data from LMArena from SiliconFlow: name from Models.dev: weights availability from OpenRouter: knowledge cutoff from NanoGPT: description from Hugging Face: release date

Advantages and trade-offs

Compared with 10 chat models, picked mainly by popularity and IQ

Trade-offs

No open weights

The weights aren't public, so you can't run it yourself.

We only make a claim when there's enough data on models of the same type. Updated 22 Nov, 06:07 UTC.Prices come from the model's maker where they sell it, otherwise their official partner, otherwise the market median. This comparison uses 7 maker's prices; 3 market median prices.

Input & output

as stated by the model's providers
Text
Text

Capabilities

Supported operations and provider-stated features

ReasoningFunction callingStructured output

Listings

 DeepSeekdeepseek-v3.1-terminusSource: GenAI Prices dataNot published by the provider$0.27cache read $0.13$0.95
 Flowdotdeepseek/deepseek-v3.1-terminusSource: FlowDot data164K$0.21$0.79
 Kilodeepseek/deepseek-v3.1-terminusSources: Kilo AI Gateway data, Models.dev data164K$0.27$1.00
 Nous Portaldeepseek/deepseek-v3.1-terminusSource: Nous Portal Inference Gateway data164K$0.27$1.00
 OpenRouterdeepseek/deepseek-v3.1-terminusSources: GenAI Prices data, Models.dev data, OpenRouter data docs164K$0.27$1.00
 Abacusdeepseek-ai/DeepSeek-V3.1-TerminusSource: Models.dev data128K$0.27$1.00
 Alibaba Cloud (China)siliconflow/deepseek-v3.1-terminusSource: Models.dev data164K$0.27$1.00
 DeepInfradeepinfra/deepseek-ai/DeepSeek-V3.1-TerminusSource: LiteLLM data164K$0.27cache read $0.216$1.00
 Gmicloudgmi/deepseek-ai/DeepSeek-V3.1-TerminusSource: LiteLLM data131K$0.27$1.00
 Novita AInovita/deepseek/deepseek-v3.1-terminus · fp8Source: LiteLLM data131K$0.27cache read $0.135$1.00
Show all 21 listings

Showing 10 of 21: the maker's own listing first, then the cheapest per 1M input tokens from each provider.

Also sold by Fireworks AI, NanoGPT, Qiniu, SiliconFlow, Siliconflow Cn, Tencent Lkeap.

2 listings without a price are hidden.

Benchmarks

Named effort levels run from lowest to highest. Select a result to see its source details.

Human preference

Arena ELO19 related benchmarks · 38 results
Source: LMArena data · Nov 2026
Benchmarknot statedThinking
Arena ELO
Creative Writing
English
Exclude Ties
Show 15 more benchmarksHide 15 benchmarks
Benchmarknot statedThinking
Hard Prompts
Hard Prompts English
Industry Business and Management and Financial Operations
Industry Entertainment and Sports and Media
Industry Life and Physical and Social Science
Industry Medicine and Healthcare
Industry Software and It Services
Industry Writing and Literature and Language
Longer Query
Multi Turn
Non English
Polish
Russian
Style-Controlled

Coding

SWE-bench-Live7 related benchmarks · 19 results
Windows
Source: SWE-bench-Live leaderboard data · Jan 2026
scaffold: win-agent, set: windows, instances: 50
MultiLang C/C++
Source: SWE-bench-Live leaderboard data · Jan 2026
By scaffold · set: ccpp · instances: 48
MultiLang C#
Source: SWE-bench-Live leaderboard data · Jan 2026
By scaffold · set: csharp · instances: 28
MultiLang Go
Source: SWE-bench-Live leaderboard data · Jan 2026
By scaffold · set: go · instances: 68
MultiLang Java
Source: SWE-bench-Live leaderboard data · Jan 2026
By scaffold · set: java · instances: 62
MultiLang Rust
Source: SWE-bench-Live leaderboard data · Jan 2026
By scaffold · set: rust · instances: 45
Show 3 more resultsHide 3 results
MultiLang TypeScript/JavaScript
Source: SWE-bench-Live leaderboard data · Jan 2026
By scaffold · set: tsjs · instances: 162

Agentic

Gaia2Source: Gaia2 (Meta ARE) data · Sep 2025
Source: OpenRouter benchmark exports · Aug 2026
deepseek-v3.1-terminus · 90-day results
Across 3 providers lowest 7.3% median 46.7% highest 76%OpenRouter's automatic routing 72.7%
Show results from 3 providers and OpenRouter's automatic routingHide results from 3 providers and OpenRouter's automatic routing
  • Novita
  • OpenRouter's automatic routing
  • SiliconFlow
  • AtlasCloud

Science

GPQA Diamond4 results
Source: OpenRouter benchmark exports · Aug 2026
deepseek-v3.1-terminus · 90-day results
Across 3 providers lowest 78.5% median 80.6% highest 82.7%OpenRouter's automatic routing 76.9%
Show results from 3 providers and OpenRouter's automatic routingHide results from 3 providers and OpenRouter's automatic routing
  • Novita
  • AtlasCloud
  • StreamLake
  • OpenRouter's automatic routing

General

Instruction following

Reasoning

Math

Multilingual

Other

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · LMArena, CC BY 4.0 (licence) · OpenRouter public benchmark leaderboard and search-lane exports, CC BY 4.0. Fractional source scores converted to percentages; immutable dated snapshots. (licence) · Shaberi v2.1, Shisa.AI, Apache-2.0. GPT-5.1-judge scores only; judges are not comparable. (licence) · SnakeBench (Greg Kamradt / ARC Prize Foundation), MIT (licence) · SWE-bench-Live (Microsoft and contributors), swe-bench-live.github.io. Resolve rates computed from the published report counts.

Also known as: deepinfra/deepseek-ai/deepseek-v3.1-terminus, deepseek-ai/deepseek-v3.1-terminus, deepseek-ai/deepseek-v3.1-terminus:thinking, deepseek-v3.1-terminus-thinking, deepseek/deepseek-v3.1-terminus, deepseek/deepseek-v3.1-terminus-thinking, deepseek/deepseek-v3.1-terminus:exacto, fireworks_ai/accounts/fireworks/models/deepseek-v3p1-terminus, gmi/deepseek-ai/deepseek-v3.1-terminus, novita/deepseek/deepseek-v3.1-terminus, pro/deepseek-ai/deepseek-v3.1-terminus, siliconflow/deepseek-v3.1-terminus

Like this one. Tap or hover any dotted value, label or icon to see where it comes from and how it is measured. open details when tapped or hovered. How this data works