R1 0528

May 2025
R1 0528
Compare in
on
—per task?
$0.39 / $1.63per 1M in / out?
26 tok/sfirst token in 1 s?
128Kcontext

Rankingin thinking mode?

Strong at

B
Reasoning#141 of 334Science#108 of 312Knowledge#144 of 318
C
Tool calling#132 of 242Long context#164 of 273Instructions#159 of 245

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperGPT OSS 120BhighOpen+14%-91%$0.092BetterGPT-5 Minihigh+9%same$0.053OpenDeepSeek V3.2 Specialethinking+16%——CheaperMiniMax M2thinkingOpen+7%-24%$0.30 / $1.20CheaperMiniMax M2.1thinkingOpen+5%-24%$0.30 / $1.20

Run it yourself

estimate
Fits on✗ 8 GB (phone, laptop)✗ 16 GB✗ 24 GB GPU✗ 48 GB GPU✗ 80 GB GPU
LicenceMITfree for commercial use

Benchmarks

all categories
Chat & Knowledge
Humanity's Last Exam15.8#141
AA Omniscience Index-27.4#145
CritPt (physics)1.4#113
AA Intelligence Index 4.3.2≈13.1#165
AA Omniscience · Accuracy30.5#75
AA Omniscience · Hallucination Rate83.4#194
AA-LCR (long context) 1.155.7#165
MMLU-Pro · official boardlab84
GPQA Diamond81.3#109
AIME 2025 · AA archived76#55
Coding
Terminal-Bench Hard (version unstated)15.9#106
LiveCodeBench · AA archived77#24
Agents & Tool Use
τ²-bench · AA36.5#134
Writing & Language
IFBench (instructions)39.6#160

FAQ

How much does R1 0528 cost?›

$0.39 per 1M input tokens and $1.63 per 1M output tokens (cheapest host: Featherless).

Is R1 0528 free to use?›

There is no free API; the cheapest paid price is $0.39 per 1M input tokens. The weights are open, so you can run it yourself.

How good is R1 0528, and what is it best at?›

#141 of 334 in Chat & Knowledge, #105 of 239 in Coding, #132 of 242 in Agents & Tool Use, #79 of 195 in Writing & Language.

Is R1 0528 better than INTELLECT 3?›

On LiveCodeBench · AA archived, R1 0528 scores 77 vs 77.7 — about the same.

How can I use R1 0528?›

Through the API of DeepInfra, Featherless, Novita, SiliconFlow, StreamLake.

What is the context window of R1 0528?›

128K tokens, with up to 33K output tokens.

When was R1 0528 released? Is there a newer version?›

Released 2025-05-28. It is the current version in its line.

Is R1 0528 open source?›

Yes, the weights are open (MIT licence).

Can I run R1 0528 locally?›

The weights are open and can be self-hosted; hardware needs are not confirmed yet.

Can R1 0528 read images?›

No, it accepts text only.

What is R1 0528?›

R1 0528 is an open-weights language model from DeepSeek, released May 2025.

What are the best alternatives to R1 0528?›

GPT OSS 120B (cheaper), GPT-5 Mini (better), DeepSeek V3.2 Speciale (open weights).