DeepSeek V4 Pro 0424
Apr 2026
Older versionReplace withDeepSeek V4 Pro 0813Newer version
GPT-6 LunaCheaper, as good
Claude Haiku 5.5Better
MiMo v2.6 ProOpen weights
DeepSeek V4 Pro 0813Newer versionGPT-6 LunaCheaper, as goodClaude Haiku 5.5BetterMiMo v2.6 ProOpen weightsDeepSeek V4 Pro 0424Compare inon
Benchmarks of DeepSeek V4 Pro 042424
BenchmarkByScoreRank
Overall1
Knowledge4
Math1
Science3
Coding4
Agents6
Health & law1
Documents2
Legacy · saturated or no longer reported2
No benchmark matches
Rankingat max effort?
Retired models are left out of the current rankings.
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperGPT-6 Lunamax+25%-44%$0.068Function calling only with none · Chat Completions+3OpenMiMo v2.6 Prothinking+52%+10%$0.13Pass back reasoning_content+3CheaperMiMo v2.6 FlashthinkingOpen+24%-49%$0.062Pass back reasoning_content+3CheaperClaude Haiku 5.5high+24%-35%$0.079Remove sampling parameters+5
CheaperAgnes 2.5 Pro Beta+16%-72%$0.10 / $0.30ModelScoreCostPer taskWhat changes
GPT-6 Lunamax+25%-44%$0.068Function calling only with none · Chat Completions+3MiMo v2.6 FlashthinkingOpen+24%-49%$0.062Pass back reasoning_content+3Claude Haiku 5.5high+24%-35%$0.079Remove sampling parameters+5Agnes 2.5 Pro Beta+16%-72%$0.10 / $0.30GPT-5.6 Lunaxhigh+13%-30%$0.085ModelScoreCostPer taskWhat changes
GPT-6 Lunahigh+8%-76%$0.029Function calling only with none · Chat Completions+3Claude Haiku 5.5medium+13%-61%$0.047Remove sampling parameters+5MiMo v2.6 FlashthinkingOpen+24%-49%$0.062Pass back reasoning_content+3GPT-5.6 Lunaxhigh+13%-30%$0.085GPT-6.1 Sollow+38%+8%$0.13Remove unsupported parameters+5ModelScoreCostPer taskWhat changes
MiMo v2.6 ProthinkingOpen+52%+10%$0.13Pass back reasoning_content+3DeepSeek V4.1 Flash 1017maxOpen+48%+116%$0.26Pass back reasoning_content · Chat Completions+4
GLM 5.3maxOpen+47%+1,550%$2.01Reasoning cannot be disabled+3
Step 5 PreviewthinkingOpen+44%+745%$1.03web_search not available · Chat Completions+2
Kimi K3maxOpen+43%+1,546%$2.00Sampling parameters are fixed+3ModelScoreCostPer taskWhat changes
DeepSeek V4.1 Flash 1017maxOpen+48%+116%$0.26Pass back reasoning_content · Chat Completions+4DeepSeek V4 Pro 0813maxOpen+18%+455%$0.67Pass back reasoning_content · Chat Completions+4DeepSeek V3.2 SpecialethinkingOpen-53%——Context: 1M → 128KDeepSeek V3.1 Terminusnon-thinkingOpen-54%—$0.27 / $0.95Context: 1M → 128KDeepSeek V3.1non-thinkingOpen-55%—$0.25 / $0.95Context: 1M → 128KR1 0528thinkingOpen-57%—$0.39 / $1.63Context: 1M → 128KDeepSeek V3 0324non-thinkingOpen-68%-83%$0.021Context: 1M → 128KDeepSeek V3non-thinkingOpen-72%-84%$0.02Context: 1M → 128KAll DeepSeek models →Run it yourself
Fits onhardware needs not verified yet
Licencenot verified
History
2 versionsBenchmarks
all categoriesChat & Knowledge
AA Omniscience Index-10.7
CritPt (physics)12.9
MLCR-AA21.1
GPQA Diamond88.8
Coding
SciCode50.8
Agents & Tool Use
τ³-Banking · AA30.1
AA Analyst Agent18.8
ITBench SRE38.3
APEX-Agents24.3
τ²-bench · AA96.2
Writing & Language
FAQ
How much does DeepSeek V4 Pro 0424 cost?›
$0.43 per 1M input tokens and $0.87 per 1M output tokens. A typical benchmark task costs about $0.12 at max effort.
Is DeepSeek V4 Pro 0424 free to use?›
There is no free API; the cheapest paid price is $0.43 per 1M input tokens. The weights are open, so you can run it yourself.
How can I use DeepSeek V4 Pro 0424?›
Through the DeepSeek API.
What is the context window of DeepSeek V4 Pro 0424?›
1M tokens.
When was DeepSeek V4 Pro 0424 released? Is there a newer version?›
Released 2026-04-24. It has been superseded by DeepSeek V4 Pro 0813.
Is DeepSeek V4 Pro 0424 open source?›
Yes, the weights are open.
Can I run DeepSeek V4 Pro 0424 locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
What is DeepSeek V4 Pro 0424?›
DeepSeek V4 Pro 0424 is an open-weights language model from DeepSeek, released Apr 2026, the successor to DeepSeek V3.2.
What are the best alternatives to DeepSeek V4 Pro 0424?›
GPT-6 Luna (cheaper), MiMo v2.6 Pro (open weights).