DeepSeek v2.5
Sep 2024
DeepSeek v2.5Compare inon
Benchmarks of DeepSeek v2.514
BenchmarkByScoreRank
No benchmark matches
—per task?
price not verifiedprice
—output speed?
—context
Rankingconfiguration unstated?
Alternatives?
ModelScoreCostPer taskWhat changes
BetterGPT-6 Lunamaxwins 64%—$0.068Function calling only with none · Chat Completions+3
OpenGLM 5.3 Flashunstatedwins 67%—$0.15 / $0.50Memory (4-bit): 141 GB → 193 GB+5BetterGPT-5.6 Lunaxhighwins 64%—$0.085BetterGPT-5.6 Terraxhighwins 66%—$0.63Prompts with >272K input tokensBetterGPT-6.1 Solmaxwins 70%—$0.72Remove unsupported parameters+5ModelScoreCostPer taskWhat changes
Nothing here yet.
ModelScoreCostPer taskWhat changes
GPT-6 Lunamaxwins 64%—$0.068Function calling only with none · Chat Completions+3GPT-5.6 Lunaxhighwins 64%—$0.085GPT-5.6 Terraxhighwins 66%—$0.63Prompts with >272K input tokensGPT-6.1 Solmaxwins 70%—$0.72Remove unsupported parameters+5
Muse Spark 1.2xhighwins 69%—$0.97Only auto tool_choice is supported+1ModelScoreCostPer taskWhat changes
GLM 5.3 FlashunstatedOpenwins 67%—$0.15 / $0.50Memory (4-bit): 141 GB → 193 GB+5
Hy3unstatedOpenwins 66%—$0.13 / $0.53Memory (4-bit): 141 GB → 179 GB
Gemma 4 31BunstatedOpenwins 65%—$0.10 / $0.34
Qwen3.5 397B A17BunstatedOpenwins 63%—$0.60 / $3.60Memory (4-bit): 141 GB → 242 GB
MiniMax M3unstatedOpenwins 63%—$0.30 / $1.20Memory (4-bit): 141 GB → 256 GB+2ModelScoreCostPer taskWhat changes
DeepSeek V4 Pro 0813highOpenwins 69%—$1.32 / $3.96Memory (4-bit): 141 GB → 990 GB+5DeepSeek V4.1 Flash 1017unstatedOpenwins 64%—$0.30 / $1.20Memory (4-bit): 141 GB → 458 GB+5DeepSeek V3.1 TerminusunstatedOpenwins 63%—$0.27 / $0.95Memory (4-bit): 141 GB → 411 GBDeepSeek V3.1thinkingOpenwins 63%—$0.25 / $0.95Memory (4-bit): 141 GB → 411 GBR1 0528unstatedOpenwins 61%—$0.39 / $1.63Memory (4-bit): 141 GB → 411 GBDeepSeek V3unstatedOpenwins 56%—$0.28 / $0.42Memory (4-bit): 141 GB → 411 GBDeepSeek Coder v2Openwins 35%——DeepSeek V3.2 SpecialethinkingOpenn/a——Memory (4-bit): 141 GB → 411 GBAll DeepSeek models →Run it yourself
estimateFits on✗ 8 GB (phone, laptop)✗ 16 GB✗ 24 GB GPU✗ 48 GB GPU✗ 80 GB GPU
Licencedeepseekcommercial use with conditions
History
5 versionsBenchmarks
all categoriesChat & Knowledge
Arena Text · overall1,323#141
Writing & Language
Arena Text · creative writing1,310#122
FAQ
How much does DeepSeek v2.5 cost?›
There is no per-unit API price: price not verified.
How good is DeepSeek v2.5, and what is it best at?›
#141 of 196 in Chat & Knowledge, #122 of 195 in Writing & Language.
Is DeepSeek v2.5 better than Step3?›
On Arena Text · creative writing, DeepSeek v2.5 scores 1,310 vs 1,310 — about the same.
When was DeepSeek v2.5 released? Is there a newer version?›
Released 2024-09-06. It is the current version in its line.
Is DeepSeek v2.5 open source?›
Yes, the weights are open (deepseek licence).
Can I run DeepSeek v2.5 locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can DeepSeek v2.5 read images?›
No, it accepts text only.
What is DeepSeek v2.5?›
DeepSeek v2.5 is an open-weights language model from DeepSeek, released Sep 2024.
What are the best alternatives to DeepSeek v2.5?›
GPT-6 Luna (better), GLM 5.3 Flash (open weights).