DeepSeek-V3.1-Terminus is described as an update to DeepSeek-V3.1 that maintains the model’s original capabilities while addressing user-reported issues, including language consistency and agent performance.
It is presented as a large hybrid reasoning model that supports both thinking and non-thinking modes.
| Developer | DeepSeek AI |
|---|---|
| Model ID | deepseek/deepseek-reasoner-v3.1-terminus |
| Release Date | August 27, 2025 |
| Context Window | 128,000 tokens (~128K) |
| Max Output Tokens | 8,000 tokens |
| Modalities | Text → Text |
| Capabilities | file input, streaming, structured (JSON schema) output, tool/function calling |
Language consistency
The model is said to reduce CN/EN mix-ups and random characters in outputs.
Agent improvements
DeepSeek states that Code Agent and Search Agent performance were further optimized.
More stable outputs
The release notes describe the model as delivering more stable and reliable outputs across benchmarks than the previous version.
Hybrid reasoning
The model supports both thinking and non-thinking modes within a single architecture.
Coding workflows
The model is positioned for coding-related tasks through improved Code Agent performance and better tool use.
Search and research tasks
The model is positioned for Search Agent workflows and structured tool calling.
Agentic workflows
The model is described as suitable for agentic workflows that benefit from stronger search and tool-use performance.
DeepSeek-V3.1
DeepSeek-V3.1-Terminus is described as an update to DeepSeek-V3.1 with improved language consistency and agent capabilities.
Sources: https://api-docs.deepseek.com/news/news250821/ · https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus · https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/maas/deepseek/deepseek-v31
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
DeepSeek V3.1 is documented as suited to advanced text generation. It is also documented for reasoning, coding, and tool use, including agentic workflows such as code agents and search agents.
It is documented as useful for code assistance. The model is also described as suited to coding and structured programming tasks.
It is documented as useful for search tasks and search-agent workflows. DeepSeek’s release note says it has stronger multi-step reasoning for complex search tasks, and the model is described as optimized for Search Agent performance.
It is documented as useful for enterprise RAG and conversational AI. One provider listing explicitly includes both as suitable uses for DeepSeek V3.1 Terminus.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "deepseek/deepseek-reasoner-v3.1-terminus", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "deepseek/deepseek-reasoner-v3.1-terminus", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"deepseek/deepseek-reasoner-v3.1-terminus","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
DeepSeek V3.1 Terminus has a documented pricing caveat: DeepSeek says its API uses cache-hit/cache-miss and peak/off-peak pricing, with off-peak rates half of peak rates, and its V3.1 release note says new pricing
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| Intelligence | 13.7 | Composite score across standardised reasoning, knowledge and problem-solving evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Math | 49.7 | Composite score across standardised mathematics evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
DeepSeek V3.1 This page | Agentic workflows and structured output | |||
| Reasoning + agents |
DeepSeek V3.1 has a 128,000 tokens context window and can return up to 8,000 tokens.
DeepSeek V3.1 takes text as input and returns text.
Use deepseek/deepseek-reasoner-v3.1-terminus as the model id. Requests go to https://api.aimlapi.com/v1/chat/completions.
DeepSeek V3.1 became available on August 28, 2025.
DeepSeek V3.1 is priced at input $0.371358 / 1M tokens, output $1.3754 / 1M tokens, cached input $0.371358 / 1M tokens.
Yes, DeepSeek V3.1 can stream responses as they are generated.
DeepSeek V3.1 was built by DeepSeek.
