DeepSeek V3.1 Terminus vs Step 3.7 Flash
Updated October 21, 2026. 17 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Neither model has a coding score yet, so the two cannot be ranked on coding.
DeepSeek V3.1 Terminus has the edge where output tokens cost 1.1x less, and where what matters is SciCode or Terminal-Bench Hard (Artificial Analysis run), where it leads.
- Shared results
- 17
- DeepSeek V3.1 Terminus only
- 27
- Step 3.7 Flash only
- 1
- Like-for-like categories
- 6 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
Step 3.7 Flash
Step 3.7 Flash leads, 19.5 against 14.8.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only DeepSeek V3.1 Terminus has a AA Agentic Index result, 6.8.
Directional only
Web design
Design Arena: websites
Step 3.7 Flash
Step 3.7 Flash leads, 1196 against 1187.
Same board, both measured
Long documents
Listed context window
Step 3.7 Flash
Step 3.7 Flash takes 262K tokens against 164K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Step 3.7 Flash
DeepSeek V3.1 Terminus $0.0008, Step 3.7 Flash $0.0008 per turn.
List prices
At a glance
- Coding index
- DeepSeek V3.1 Terminus not benchmarked, Step 3.7 Flash not benchmarked.
- Widest category
- Agentic work, 61.4 points apart on average over 1 shared benchmark.
- Widest single gap
- τ²-Bench Telecom: DeepSeek V3.1 Terminus 37.1% against Step 3.7 Flash 98.5%, 61.4 points apart on a 0 to 100 scale.
- Output price
- DeepSeek V3.1 Terminus $1.0142 against Step 3.7 Flash $1.15 per million output tokens, 1.1x apart.
- Context window
- DeepSeek V3.1 Terminus takes 164K tokens, Step 3.7 Flash 262K.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | DeepSeek V3.1 Terminus | Bars | Step 3.7 Flash | Ties |
|---|---|---|---|---|
| Coding | 2 | 2 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 0 | 1 | ||
| Reasoning | 0 | 2 | ||
| Web design | 2 | 5 | ||
| General | 0 | 2 | ||
| All shared boards | 4 | 13 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index22
| Benchmark | DeepSeek V3.1 Terminus | Step 3.7 Flash | Leader |
|---|---|---|---|
| AA Coding Index | 43.5 | 39.6 | DeepSeek V3.1 Terminus |
| LiveCodeBench (Artificial Analysis run) | 79.8% | Not measured | Only DeepSeek V3.1 Terminus measured |
| SciCode | 38.0% | 43.9% | Step 3.7 Flash |
| Terminal-Bench Hard (Artificial Analysis run) | 31.8% | 35.6% | Step 3.7 Flash |
| ALE-Bench | 745.2 | 694.1 | DeepSeek V3.1 Terminus |
| LMArena Text: coding | 1425 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: software and IT | 1426 | Not measured | Only DeepSeek V3.1 Terminus measured |
| Neither model has a result on the other 31 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V3.1 Terminus | Step 3.7 Flash | Leader |
|---|---|---|---|
| AA Intelligence Index | 14.8 | 19.5 | Step 3.7 Flash |
| Neither model has a result on the other 1 board here | |||
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V3.1 Terminus | Step 3.7 Flash | Leader |
|---|---|---|---|
| AA Agentic Index | 6.8 | Not measured | Only DeepSeek V3.1 Terminus measured |
| τ²-Bench Telecom | 37.1% | 98.5% | Step 3.7 Flash |
| Neither model has a result on the other 37 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | DeepSeek V3.1 Terminus | Step 3.7 Flash | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | 79.2% | 80.9% | Step 3.7 Flash |
| Humanity's Last Exam (Artificial Analysis run) | 16.4% | 21.4% | Step 3.7 Flash |
| LMArena Text: hard prompts | 1426 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: hard prompts in English | 1432 | Not measured | Only DeepSeek V3.1 Terminus measured |
| Neither model has a result on the other 17 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index3 boards, none shared
| Benchmark | DeepSeek V3.1 Terminus | Step 3.7 Flash | Leader |
|---|---|---|---|
| AA Math Index | 89.7 | Not measured | Only DeepSeek V3.1 Terminus measured |
| AIME 2025 (Artificial Analysis run) | 89.7% | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: math | 1402 | Not measured | Only DeepSeek V3.1 Terminus measured |
| Neither model has a result on the other 10 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index5 boards, none shared
| Benchmark | DeepSeek V3.1 Terminus | Step 3.7 Flash | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | 85.1% | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: science | 1451 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: medicine and healthcare | 1468 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: legal and government | 1457 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: business and finance | 1414 | Not measured | Only DeepSeek V3.1 Terminus measured |
| Neither model has a result on the other 12 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index25
| Benchmark | DeepSeek V3.1 Terminus | Step 3.7 Flash | Leader |
|---|---|---|---|
| Design Arena: websites | 1187 | 1196 | Step 3.7 Flash |
| Design Arena: UI components | 1182 | 1179 | DeepSeek V3.1 Terminus |
| Design Arena: game dev | 1131 | 1158 | Step 3.7 Flash |
| Design Arena: data viz | 1167 | 1172 | Step 3.7 Flash |
| Design Arena: 3D | 1141 | 1129 | DeepSeek V3.1 Terminus |
| Design Arena: all code categories | 1175 | 1182 | Step 3.7 Flash |
| Design Arena: SVG | 1065 | 1085 | Step 3.7 Flash |
| Design Arena: ASCII art | Not measured | 1152 | Only Step 3.7 Flash measured |
| Neither model has a result on the other 1 board here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | DeepSeek V3.1 Terminus | Step 3.7 Flash | Leader |
|---|---|---|---|
| IFBench | 57.0% | 67.3% | Step 3.7 Flash |
| Long context reasoning (AA-LCR) | 69.3% | 73.7% | Step 3.7 Flash |
| DTBench | 81.3% | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMCA | 28.6% | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text | 1419 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: instruction following | 1404 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: creative writing | 1403 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: English | 1429 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: non-English | 1408 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: Russian | 1435 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: Polish | 1427 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: multi-turn | 1411 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: longer queries | 1422 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: writing and language | 1409 | Not measured | Only DeepSeek V3.1 Terminus measured |
| LMArena Text: entertainment and media | 1391 | Not measured | Only DeepSeek V3.1 Terminus measured |
| Neither model has a result on the other 31 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | DeepSeek V3.1 Terminus | Step 3.7 Flash | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
Step 3.7 Flash costs less
Repository review
50K input + 3K output tokens
Step 3.7 Flash costs less
Agent loop
200K input + 10K output tokens
Step 3.7 Flash costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | DeepSeek V3.1 Terminus | StepFun Step 3.7 Flash |
|---|---|---|
| Maker | DeepSeek | StepFun |
| Weights | Open weights | Open weights |
| Released | September 22, 2025 | May 28, 2026 |
| Context window | 164K | 262K |
| Input $/Mtok | $0.3063 | $0.20 |
| Output $/Mtok | $1.0142 | $1.15 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 0 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is DeepSeek V3.1 Terminus better than Step 3.7 Flash for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. DeepSeek V3.1 Terminus has the edge where output tokens cost 1.1x less, and where what matters is SciCode or Terminal-Bench Hard (Artificial Analysis run), where it leads.
Which is cheaper, DeepSeek V3.1 Terminus or Step 3.7 Flash?
DeepSeek V3.1 Terminus costs $0.3063 per million input tokens and $1.0142 per million output tokens; Step 3.7 Flash costs $0.20 and $1.15. Cost per solved task is known for neither of them.
Which has the bigger context window, DeepSeek V3.1 Terminus or Step 3.7 Flash?
Step 3.7 Flash, with 262K tokens against 164K.
How many benchmarks is the DeepSeek V3.1 Terminus vs Step 3.7 Flash comparison based on?
17 benchmarks have a result for both models as of October 21, 2026. DeepSeek V3.1 Terminus's coding index rests on 0 independent benchmarks and Step 3.7 Flash's on 0. Where only one model has a result, the page says so instead of naming a winner.