DeepSeek model product
DeepSeek-V2.5
5 is an upgraded version that combines DeepSeek-V2-Chat and DeepSeek-Coder-V2-Instruct, integrating general and coding abilities.
Updated Aug 17, 2026. Default version: DeepSeek-V2.5
LLMBoard Score13.8DeepSeek-V2.5
Coverage100%11 benchmark families
Context window8.2KTokens
Official input priceN/AOfficial price unavailable
DeepSeek-V2.5 Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
DeepSeek-V2.5 Benchmark Results
Benchmark scores for DeepSeek-V2.5.
Sort by
Benchmark | Score | Rank | Participants | Percentile | Evidence | Evaluated |
|---|
| BenchmarkAider | Score72.2% | Rank01 | Participants4 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkDS-Arena-Code | Score63.1% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkDS-FIM-Eval | Score78.3% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkLiveCodeBench(01-09) | Score41.8% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkAlignBench | Score80.4% | Rank02 | Participants4 | Percentile66.7% | EvidenceC | Evaluated |
| BenchmarkHumanEval-Mul | Score73.8% | Rank02 | Participants2 | Percentile0.0% | EvidenceC | Evaluated |
| BenchmarkAlpacaEval 2.0 | Score50.5% | Rank03 | Participants4 | Percentile33.3% | EvidenceC | Evaluated |
| BenchmarkMT-Bench | Score90.2% | Rank03 | Participants12 | Percentile81.8% | EvidenceC | Evaluated |
| BenchmarkBBH | Score84.3% | Rank05 | Participants12 | Percentile63.6% | EvidenceC | Evaluated |
| BenchmarkArena Hard | Score76.2% | Rank08 | Participants26 | Percentile72.0% | EvidenceC | Evaluated |
| BenchmarkGSM8k | Score95.1% | Rank11 | Participants48 | Percentile78.7% | EvidenceC | Evaluated |
| BenchmarkHumanEval | Score89.0% | Rank16 | Participants66 | Percentile76.9% | EvidenceC | Evaluated |
| BenchmarkMATH | Score74.7% | Rank29 | Participants71 | Percentile60.0% | EvidenceC | Evaluated |
| BenchmarkMMLU | Score80.4% | Rank60 | Participants101 | Percentile41.0% | EvidenceC | Evaluated |
| BenchmarkSWE-Bench Verified | Score16.8% | Rank110 | Participants111 | Percentile0.9% | EvidenceC | Evaluated |
DeepSeek-V2.5 Arena Results
Preference and agent-evaluation results for the default version.
Sort by
Arena | Category | Rank | Rating / score | Votes | Observations | Result date |
|---|
| Arenatext | Categoryoverall | Rank227 | Rating / score1294.1 | Votes6,795 | ObservationsN/A | Result date |
| Arenatext style control | Categoryoverall | Rank231 | Rating / score1323.5 | Votes6,795 | ObservationsN/A | Result date |
| Arenatext style control | Categoryoverall | Rank252 | Rating / score1307.2 | Votes24,572 | ObservationsN/A | Result date |
| Arenatext | Categoryoverall | Rank253 | Rating / score1271.4 | Votes24,572 | ObservationsN/A | Result date |
DeepSeek-V2.5 Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- N/A
- Tracked offerings
- 0
No provider prices
The default version has no current input or output token prices.
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
DeepSeek-V2.5 Runtime Performance
Provider-specific output speed and catalog latency for DeepSeek-V2.5. Runtime does not affect the capability score.
Sort by
Provider | Output Speed | Catalog Latency | Max Input | Max Output | Updated |
|---|
| ProviderDeepSeek | Output Speed100 tok/s | Catalog Latency0.5 s | Max Input8.2K | Max Output8.2K | Updated |
| ProviderHyperbolic | Output Speed100 tok/s | Catalog Latency0.5 s | Max Input8.2K | Max Output8.2K | Updated |
| ProviderDeepInfra | Output Speed63 tok/s | Catalog Latency0.5 s | Max Input8.2K | Max Output8.2K | Updated |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
DeepSeek-V2.5 Specifications
Technical details for the model's default version.
- Version
- DeepSeek-V2.5
- Released
- May 8, 2024
- Knowledge cutoff
- Unknown
- Parameters
- 236B
- Context window
- 8.2K
- Max output
- 8.2K
- Inputs
- text
- Outputs
- text
- Open weights
- No
- License
- deepseek
DeepSeek-V2.5 Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
Sort by
Version | Released | LLMBoard | Parameters | Context | Max output | Open weights | License |
|---|
| VersionDeepSeek-V2.5 | Released | LLMBoard13.8 | Parameters236B | Context8.2K | Max output8.2K | Open weightsNo | Licensedeepseek |
DeepSeek-V2.5 vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to DeepSeek-V2.5
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
DE
DeepSeek-R1-Distill-Qwen
DeepSeek
DE
DeepSeek-R1-Distill-Llama
DeepSeek
DE
DeepSeek-V3
DeepSeek
DE
DeepSeek-VL2
DeepSeek
DE
DeepSeek-V3.1
DeepSeek
DE
DeepSeek-R1
DeepSeek
What is DeepSeek-V2.5?
Key information about DeepSeek-V2.5 and its available data.
5 is an upgraded version that combines DeepSeek-V2-Chat and DeepSeek-Coder-V2-Instruct, integrating general and coding abilities. It better aligns with human preferences and has been optimized in various aspects, including writing and instruction following.
Data as of 2026-08-17.
FAQ
Common questions about DeepSeek-V2.5.
When was DeepSeek-V2.5 released?
DeepSeek-V2.5's default version was released on May 8, 2024.
How much does DeepSeek-V2.5 cost?
No official standard PAYG price is currently available for DeepSeek-V2.5.
Who created DeepSeek-V2.5?
DeepSeek-V2.5 was created by DeepSeek.
What is the context window for DeepSeek-V2.5?
The default version has a 8.2K token context window.
Is DeepSeek-V2.5 open weight?
No. The default version is not marked as having publicly available weights.
How many API providers offer DeepSeek-V2.5?
No provider offering is currently linked to the default version.
What models should I compare DeepSeek-V2.5 with?
Nearby ranked alternatives include QvQ 72B, Qwen2.5 32B, GPT-4-Turbo.