Gemini 4 Argon
Claude Opus 5.5
Muse Wonder
MiMo-V2.6-Pro
Qwen3.8-Max
Kimi K3
GLM-5.3
GPT-5.5
ERNIE 5.1
Grok 4.20 Beta
amazon-nova-experimental-chat
Dola Seed 2.0 Pro
NVIDIA-Nemotron-3-Ultra-550B-A55B
Gemma 4 31B
Step 5 Preview
DeepSeek-V4-Flash
DeepSeek-V4-Flash is an open-weight model from DeepSeek, released Aug 1, 2026. It ranks #91 of 351 on the LMArena text leaderboard with a rating of 1432. It costs $0.042 per million input tokens and $0.18 per million output tokens, with a 1M-token context window.
- Developer
- DeepSeek
- Released
- Aug 1, 2026
- Context window
- 1M
- Price in / out, per 1M tokens
- $0.042 / $0.18
- Weights
- Open · MIT
- Inputs
- Text
- Reasoning
- Yes
- Tool use
- Yes
The rating LMArena published at each update since Apr 23, 2026. Ratings move as votes come in and new models join.
Benchmarks
| Benchmark | Score bar | Accuracy | Rank | Cost / test |
|---|---|---|---|---|
| Terminal-Bench 4.0Coding | 18.69% | 32 / 55 | $0.77 | |
| SWE-benchCoding | 88.80% | 11 / 85 | $0.01 | |
| Vibe Code BenchCoding | 74.74% | 37 / 115 | $0.91 | |
| LiveCodeBenchCoding | 87.26% | 17 / 143 | $0.44 / $1.32 | |
| Tax Agent BenchFinance | 28.07% | 33 / 76 | $0.16 | |
| Legal Research BenchLaw | 30.29% | 50 / 85 | $0.28 | |
| MedCodeMedicine | 41.41% | 53 / 107 | $0.44 / $1.32 | |
| GPQA DiamondReasoning | 89.90% | 27 / 138 | $0.44 / $1.32 | |
| MMLU-ProReasoning | 86.21% | 47 / 138 | $0.44 / $1.32 | |
| Terminal-Bench ScienceReasoning | 4.29% | 27 / 42 | $1.10 |
Vals AI, as of Oct 24, 2026.
LMArena, by category
| Leaderboard | Rating | Rank | Votes |
|---|---|---|---|
| TextLMArena | 1432 | 91 / 351 | 52K |
| CodingLMArena | 1457 | 89 / 346 | 15.3K |
| MathLMArena | 1426 | 110 / 339 | 2.7K |
| Creative writingLMArena | 1403 | 93 / 349 | 8.6K |
| Instruction followingLMArena | 1421 | 85 / 351 | 18K |
| Hard promptsLMArena | 1444 | 88 / 351 | 34.8K |
| Multi-turnLMArena | 1448 | 69 / 349 | 9.3K |
| Long queriesLMArena | 1434 | 88 / 330 | 23.4K |
| ExpertLMArena | 1441 | 97 / 305 | 5.4K |
| Legal and governmentIndustries | 1439 | 89 / 325 | 4.1K |
| Medicine and healthcareIndustries | 1448 | 79 / 325 | 3.7K |
| Business and financeIndustries | 1425 | 93 / 344 | 10.4K |
| ChineseLanguages | 1467 | 91 / 328 | 3K |
| SpanishLanguages | 1435 | 90 / 249 | 1.6K |
| GermanLanguages | 1419 | 98 / 261 | 839 |
| FrenchLanguages | 1440 | 106 / 252 | 2K |
| JapaneseLanguages | 1406 | 71 / 236 | 640 |
| WebDevLMArena | 1582 | 38 / 124 | 4.9K |
LMArena, CC BY 4.0, as of Oct 22, 2026. Where a model is rated at several effort levels, its best is shown (DeepSeek-V4-Flash, DeepSeek-V4-Flash Preview (High), DeepSeek-V4-Flash (High)).
Compare DeepSeek-V4-Flash
DeepSeek
A research lab known for strong open-weight reasoning and coding models trained at low cost. Founded 2023, based in Hangzhou. deepseek.com
Questions
Who makes DeepSeek-V4-Flash?
DeepSeek-V4-Flash is made by DeepSeek, founded 2023 and based in Hangzhou.
How much does DeepSeek-V4-Flash cost?
$0.042 per million input tokens and $0.18 per million output tokens, at its provider’s list price.
What is DeepSeek-V4-Flash’s context window?
1M tokens (1,048,576).
How does DeepSeek-V4-Flash rank?
#91 of 351 on LMArena’s text leaderboard, rated 1432 (95% interval 1428–1436) from 51,982 votes, as of Oct 22, 2026.
Is DeepSeek-V4-Flash open source?
DeepSeek-V4-Flash has open weights under the MIT license.