DeepSeek V3.1 Terminus
DeepSeek · released Sep 22, 2025 · deepseek-ai/DeepSeek-V3.1-Terminus
- Type
- Open weightsMIT License
- Params
- 685B
- Context
- 164K
37B active per word · about 123K words of context
Our take
Written Sep 2, 2026DeepSeek V3.1 Terminus is a large downloadable text model with a permissive MIT licence and 37 billion active parameters per token. It scores highest on coding among its measured variants and offers wide provider choice, though it handles text only.
Pick this for coding workloads where its measured Arena score leads its own variants, or for long-context text processing up to 163,840 tokens. Use it if you want open weights with a genuinely permissive licence and real provider competition. Skip it if you need image or video input, if instruction-following or mathematics quality is paramount, or if you want the cheapest option and cannot tolerate a fourfold throughput gap to the fastest host.
The case for it
- Highest measured coding score among its own benchmark variants, leading the next nearest by 16.9 points.
- MIT licence permits commercial use, modification and redistribution without restriction.
- Thirteen current hosted offers, with strong price competition at the entry tier.
- Throughput up to 44 tokens per second available for latency-sensitive workloads.
The case against it
- Instruction following and mathematics are its weakest measured areas, sitting 30.9 and 20.4 points below its overall score respectively.
- Fastest throughput costs markedly more than the cheapest option — a 37% input premium for the 44 tokens-per-second tier.
- Text-only; no image, audio or video input.
How good is it?
EverydayGeneral questions and everyday reasoning
Arena Text (overall)84th of 168 · 1415
Also on this board: 1417 (Sep 25, 2026). Read the pair, not the higher one.
CodingWriting and fixing code on its own
Arena Coding100th of 168 · 1440
Arena Coding is the only board that has scored it for this.
Also on this board: 1460 (Sep 25, 2026). Read the pair, not the higher one.
AgenticPlanning, calling tools, staying on task
Not yet scored on Arena Agent.
WritingDrafting and rewriting prose
Arena Creative Writing62nd of 168 · 1405
Arena Creative Writing is the only board that has scored it for this.
Also on this board: 1388 (Sep 25, 2026). Read the pair, not the higher one.
These tests check whether a model follows instructions — a precondition for all the work above, but not a measure of how well that work is done.
Every published score for this model6 scoresEvery figure we hold, from 6 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Can you run it yourself?
GeForce RTX 4090 · 24 GB
Too large for this card. The weights do not fit even with part of them offloaded to system memory.
Apple M1 Pro (16-core GPU) · 32 GB
Too large for this card. The weights do not fit even with part of them offloaded to system memory.
Radeon RX 7900 XT · 20 GB
Too large for this card. The weights do not fit even with part of them offloaded to system memory.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first.
Check against your own machine → · Where to rent it hosted →
Or rent it from someone else
Prices checked between 1 hour and 4 days ago — each listing carries its own date.
- per 1M tokens
- $0.30 in / $1.00 out
- Context served
- 164K
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| OpenRouterOpenRouter's own listing | $0.30 / $1.00checked 1 hour ago | 164K | not measured | Unknown | Unknown | Unknown |
| Novita AIDirect | $0.27 / $1.00checked 4 days ago | 131K | not measured | Unknown | Unknown | Unknown |
| SiliconFlowfp8Through OpenRouter | $0.27 / $1.00checked 1 hour ago | 164K147K max reply | 17 tok/s | No | No | Confirmed |
| AtlasCloudfp8Through OpenRouter | $0.30 / $1.00checked 1 hour ago | 131K66K max reply | 52 tok/s | No | Yesunknown period | Unknown |
| StreamLakeThrough OpenRouter | $0.34 / $1.03checked 1 hour ago | 128K32K max reply | 57 tok/s | No | Yesunknown period | Unknown |
Novita AIAcross the 5 listings we hold: 3 say they do not train on prompts, 0 say they do and 2 do not say. 1 appears in the zero-retention registry we check; the rest are unknown to us.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| OpenRouterOpenRouter's own listing | ✓ | ✓ | ✓ |
| Novita AIDirect | |||
| SiliconFlowfp8Through OpenRouter | ✓ | ✓ | ✓ |
| AtlasCloudfp8Through OpenRouter | ✓ | ✓ | ✓ |
| StreamLakeThrough OpenRouter | ✗ | ✗ | ✓ |
Tool calling: 3 of 5 listings say yes, 1 says no, 1 publishes no parameter list. JSON output: 3 of 5 listings say yes, 1 says no, 1 publishes no parameter list. Strict schema: 4 of 5 listings say yes, 1 publishes no parameter list.
When we formed this view
Recent changes
What moved
output +5% ($0.95 → $1.00 per 1M tokens), cache read +4% ($0.130 → $0.135 per 1M tokens)What moved
first indexed by our pipelineEach date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
- 1 of 5 listings publishes no parameter list, so what its API accepts is unknown to us.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 2 of 5 listings do not say whether they train on prompts.
- We hold no batch or off-peak rate for any of its listings.
- We hold a decode speed for it, but no prompt-processing (prefill) figure, so how long the input side of a job takes is unknown to us.
Licence and identifiers
What the licence allowsMIT License, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
MIT License
Fully permissive: do anything with attribution. No patent grant, unlike Apache-2.0.
Identifiers
- Hugging Face
- deepseek-ai/DeepSeek-V3.1-Terminus
- Architecture
- Mixture of experts
- Takes in, gives back
- Text in, text out
- Catalogue slug
- deepseek-deepseek-v3-1-terminus