Model comparison
DeepSeek-V3.1-Terminus vs DeepSeek V4 Flash
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 43.1 on the Noometry Index.
Last verified . 16 shared benchmarks.
DeepSeekSummary
- They share 16 benchmarks with published results for both. DeepSeek-V3.1-Terminus scores higher in 0 categories and DeepSeek V4 Flash in 7 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Flash leads 53.7 to 26.4.
- The biggest single-benchmark swing is CritPt: 1.7% for DeepSeek-V3.1-Terminus and 16.6% for DeepSeek V4 Flash.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $0.27 / $1 for DeepSeek-V3.1-Terminus.
- DeepSeek V4 Flash accepts more context: 1M tokens versus 164K.
Side by side
| DeepSeek-V3.1-Terminus | DeepSeek V4 Flash | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 43.1 | 53.6 |
| Released | 2025-09-22 | 2026-04-24 |
| Weights | Open | Open |
| Context window | 164K | 1M |
| Max output | 147K | 393K |
| Input $ / M tokens | $0.27 | $0.15 |
| Output $ / M tokens | $1 | $0.60 |
| Results tracked | 16 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Flash leads
DeepSeek-V3.1-Terminus: 42.0 (#113), DeepSeek V4 Flash: 47.9 (#59)
| Benchmark | DeepSeek-V3.1-Terminus | DeepSeek V4 Flash |
|---|---|---|
| SciCode | 40.6% | 49.9% |
| LMArena Coding | 1426 | 1457 |
| ALE-Bench | 745.17 | 1,306 |
| FrontierCode | — | 18.8% |
| LMArena WebDev | — | 1582 |
| WeirdML | — | 63% |
Reasoning DeepSeek V4 Flash leads
DeepSeek-V3.1-Terminus: 26.4 (#133), DeepSeek V4 Flash: 53.7 (#30)
| Benchmark | DeepSeek-V3.1-Terminus | DeepSeek V4 Flash |
|---|---|---|
| Kagi LLM Benchmark | 57.4% | 52.2% |
| CritPt | 1.7% | 16.6% |
| LMArena Hard Prompts | 1426 | 1444 |
| DTBench | 81.3% | 90.9% |
| LMCA | 28.6% | 41.7% |
| ARC-AGI-2 | — | 61.4% |
| SimpleBench | — | 61.1% |
| NYT Connections (extended) | — | 89.6% |
| ARC-AGI-1 | — | 89% |
| Chess Puzzles | — | 33% |
| Mystery Game Puzzles | — | 34% |
| Epoch Capabilities Index | — | 154.49 |
Math DeepSeek V4 Flash leads
DeepSeek-V3.1-Terminus: 38.5 (#137), DeepSeek V4 Flash: 60.3 (#37)
| Benchmark | DeepSeek-V3.1-Terminus | DeepSeek V4 Flash |
|---|---|---|
| LMArena Math | 1402 | 1427 |
| FrontierMath (Tiers 1-3) | — | 57.5% |
| FrontierMath Tier 4 | — | 24.4% |
| MathArena Final-Answer Competitions | — | 76.5% |
| OTIS Mock AIME 2024-2025 | — | 94.4% |
| ProofBench | — | 56% |
Knowledge Not comparable
DeepSeek-V3.1-Terminus: —, DeepSeek V4 Flash: 55.4 (#48)
| Benchmark | DeepSeek-V3.1-Terminus | DeepSeek V4 Flash |
|---|---|---|
| GPQA Diamond | — | 91% |
| SimpleQA Verified | — | 33.6% |
| LMArena Expert | — | 1441 |
Multilingual Too close to call
DeepSeek-V3.1-Terminus: 52.1 (#92), DeepSeek V4 Flash: 53.0 (#72)
| Benchmark | DeepSeek-V3.1-Terminus | DeepSeek V4 Flash |
|---|---|---|
| LMArena Non-English | 1407 | 1420 |
| LMArena Russian | 1436 | 1428 |
| LMArena Chinese | — | 1468 |
| LMArena French | — | 1439 |
| LMArena German | — | 1418 |
| LMArena Japanese | — | 1406 |
| LMArena Korean | — | 1384 |
| LMArena Spanish | — | 1436 |
Instruction Following Too close to call
DeepSeek-V3.1-Terminus: 74.0 (#106), DeepSeek V4 Flash: 74.9 (#81)
| Benchmark | DeepSeek-V3.1-Terminus | DeepSeek V4 Flash |
|---|---|---|
| LMArena Instruction Following | 1404 | 1421 |
Long Context Too close to call
DeepSeek-V3.1-Terminus: 43.4 (#97), DeepSeek V4 Flash: 43.8 (#85)
| Benchmark | DeepSeek-V3.1-Terminus | DeepSeek V4 Flash |
|---|---|---|
| LMArena Longer Query | 1421 | 1434 |
Writing & Preference DeepSeek V4 Flash leads
DeepSeek-V3.1-Terminus: 61.0 (#92), DeepSeek V4 Flash: 63.8 (#61)
| Benchmark | DeepSeek-V3.1-Terminus | DeepSeek V4 Flash |
|---|---|---|
| LMArena Text | 1419 | 1432 |
| LMArena Creative Writing | 1403 | 1403 |
| LMArena Multi-Turn | 1411 | 1449 |
| EQ-Bench Creative Writing | — | 1559 |
Frequently asked questions
Is DeepSeek-V3.1-Terminus better than DeepSeek V4 Flash?
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 43.1 on the Noometry Index.
Which is cheaper, DeepSeek-V3.1-Terminus or DeepSeek V4 Flash?
DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; DeepSeek-V3.1-Terminus lists at $0.27 and $1.
Is DeepSeek-V3.1-Terminus or DeepSeek V4 Flash better for coding?
DeepSeek V4 Flash scores higher on coding benchmarks: 47.9 versus 42.0 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4 Flash does, with 1M tokens against 164K.
How many benchmarks do DeepSeek-V3.1-Terminus and DeepSeek V4 Flash share?
16 benchmarks have published results for both models. DeepSeek-V3.1-Terminus has 16 scored results on Noometry and DeepSeek V4 Flash has 41.