DeepSeek model product
DeepSeek-V2.5 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.
Updated Aug 10, 2026. Default version: DeepSeek-V2.5
Structured fields from the published default version.
This profile uses the latest version under this unique model that has a calculated LLMBoard score. Arena and price are excluded.
Published benchmark records for the scored version DeepSeek-V2.5.
| Aider | 0.7 | 1 | 4 | 100.0% | C | |
| DS-Arena-Code | 0.6 | 1 | 1 | 100.0% | C | |
| DS-FIM-Eval | 0.8 | 1 | 1 | 100.0% | C | |
| LiveCodeBench(01-09) | 0.4 | 1 | 1 | 100.0% | C | |
| AlignBench | 0.8 | 2 | 4 | 66.7% | C | |
| HumanEval-Mul | 0.7 | 2 | 2 | 0.0% | C | |
| AlpacaEval 2.0 | 0.5 | 3 | 4 | 33.3% | C | |
| MT-Bench | 0.9 | 3 | 12 | 81.8% | C | |
| BBH | 0.8 | 5 | 12 | 63.6% | C | |
| Arena Hard | 0.8 | 8 | 26 | 72.0% | C | |
| GSM8k | 1.0 | 11 | 48 | 78.7% | C | |
| HumanEval | 0.9 | 18 | 76 | 77.3% | C | |
| MATH | 0.7 | 29 | 71 | 60.0% | C | |
| MMLU | 0.8 | 60 | 100 | 40.4% | C | |
| SWE-Bench Verified | 0.2 | 103 | 104 | 1.0% | C |
Preference and agent-evaluation signals from published Arena datasets.
| text | japanese | 187 | 1190.5 | 479 | N/A | |
| text style control | japanese | 188 | 1227.6 | 479 | N/A | |
| text | french | 190 | 1287.7 | 222 | N/A | |
| text | korean | 190 | 1209.6 | 344 | N/A | |
| text style control | korean | 190 | 1254.8 | 344 | N/A | |
| text style control | french | 202 | 1324.7 | 222 | N/A | |
| text | spanish | 213 | 1247.2 | 293 | N/A | |
| text | chinese | 217 | 1280.4 | 2,242 | N/A | |
| text style control | chinese | 219 | 1329.6 | 2,242 | N/A | |
| text style control | coding | 219 | 1368.4 | 4,252 | N/A | |
| text style control | spanish | 220 | 1283.0 | 293 | N/A | |
| text | german | 221 | 1226.3 | 563 | N/A | |
| text style control | german | 223 | 1261.1 | 563 | N/A | |
| text | coding | 227 | 1301.8 | 4,252 | N/A | |
| text style control | industry legal and government | 229 | 1330.2 | 1,518 | N/A | |
| text | industry software and it services | 232 | 1296.1 | 6,692 | N/A | |
| text style control | industry software and it services | 232 | 1349.8 | 6,692 | N/A | |
| text | expert | 233 | 1239.0 | 1,523 | N/A | |
| text | industry mathematical | 233 | 1273.5 | 3,039 | N/A | |
| text | industry legal and government | 234 | 1282.2 | 1,518 | N/A | |
| text | industry medicine and healthcare | 235 | 1258.5 | 1,214 | N/A | |
| text | math | 235 | 1270.7 | 3,649 | N/A | |
| text | hard prompts | 236 | 1270.7 | 6,869 | N/A | |
| text style control | expert | 236 | 1294.8 | 1,523 | N/A | |
| text style control | math | 237 | 1287.5 | 3,649 | N/A | |
| text style control | hard prompts | 238 | 1321.5 | 6,869 | N/A | |
| text style control | industry mathematical | 238 | 1295.7 | 3,039 | N/A | |
| text | hard prompts english | 240 | 1281.5 | 4,055 | N/A | |
| text | longer query | 241 | 1273.3 | 3,499 | N/A | |
| text style control | hard prompts english | 241 | 1330.5 | 4,055 | N/A | |
| text style control | non english | 241 | 1290.0 | 11,727 | N/A | |
| text | industry business and management and financial operations | 242 | 1247.9 | 2,825 | N/A | |
| text | non english | 243 | 1253.0 | 11,727 | N/A | |
| text | russian | 243 | 1251.9 | 3,519 | N/A | |
| text style control | longer query | 243 | 1314.2 | 3,499 | N/A | |
| text style control | industry business and management and financial operations | 244 | 1300.0 | 2,825 | N/A | |
| text style control | russian | 244 | 1290.4 | 3,519 | N/A | |
| text style control | industry medicine and healthcare | 245 | 1313.6 | 1,214 | N/A | |
| text | industry life and physical and social science | 246 | 1273.6 | 4,318 | N/A | |
| text | instruction following | 246 | 1249.5 | 10,175 | N/A | |
| text style control | instruction following | 247 | 1291.5 | 10,175 | N/A | |
| text | exclude ties | 248 | 1193.2 | 15,337 | N/A | |
| text style control | overall | 248 | 1307.2 | 24,572 | N/A | |
| text | overall | 249 | 1271.4 | 24,572 | N/A | |
| text style control | industry life and physical and social science | 249 | 1320.8 | 4,318 | N/A | |
| text style control | multi turn | 249 | 1292.6 | 4,473 | N/A | |
| text | english | 250 | 1284.4 | 12,845 | N/A | |
| text | multi turn | 250 | 1260.7 | 4,473 | N/A | |
| text style control | industry entertainment and sports and media | 250 | 1264.0 | 3,664 | N/A | |
| text style control | industry writing and literature and language | 251 | 1283.5 | 6,613 | N/A | |
| text | industry entertainment and sports and media | 252 | 1230.5 | 3,664 | N/A | |
| text | creative writing | 253 | 1237.1 | 3,450 | N/A | |
| text style control | creative writing | 253 | 1264.9 | 3,450 | N/A | |
| text style control | exclude ties | 253 | 1243.6 | 15,337 | N/A | |
| text style control | english | 258 | 1318.5 | 12,845 | N/A | |
| text | industry writing and literature and language | 259 | 1246.8 | 6,613 | N/A |
Official vendor API PAYG pricing is summarized first. The table then lists individual provider offerings without treating their minimum as the official price.
The default version has no provider offering with 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.
All published versions linked to this unique model. The score columns identify the version used by the current overall ranking.
| DeepSeek-V2.5 | 69.5 | 236B | 8.2K | 8.2K | No | deepseek |
A concise description based on the published model registry.
DeepSeek-V2.5 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.
Use the benchmark, Arena and pricing sections above as separate evidence. A missing field means the current data snapshot does not support that claim.
Data snapshot: 2026-08-07. Editorial model content is not available in the backend.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest published LLMBoard score.
Common questions about DeepSeek-V2.5.
DeepSeek-V2.5's default version was released on May 8, 2024.
No official standard PAYG price is currently available for DeepSeek-V2.5.
DeepSeek-V2.5 is published under DeepSeek in the model registry.
The default version has a 8.2K token context window.
No. The default version is not marked as having publicly available weights.
No published provider offering is currently linked to the default version.
Nearby ranked alternatives include Kimi K2.5, GPT-5, Llama 3.1 405B.