LIVE
Introducing Cosmos 3 Edge20/07/26 · Hugging Face|Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling20/07/26 · Anthropic|At SIGGRAPH, NVIDIA Advances Graphics and Simulation With Agentic and Physical AI20/07/26 · NVIDIA|China's open-weights AI strategy is winning20/07/26|Bristol Myers Squibb Building Life Science Industry’s Most Advanced AI Factory on NVIDIA Vera Rubin20/07/26 · NVIDIA|Safety and alignment in an era of long-horizon models20/07/26 · OpenAI|Claude Fable produced a counterexample to the Jacobian Conjecture20/07/26 · Anthropic|Apply for Anthropic’s AI for Science rare disease research grants20/07/26 · Anthropic|AI advice made people less accurate but more confident – sudy19/07/26|Vincentwei1021/video-shotcraft: AI video skill for Claude Code & Codex — cinematic product videos with Remotion: 106 shot recipe cards, 161 motion previews, a production-ready template19/07/26 · Anthropic|Claude Code uses Bun written in Rust now19/07/26 · Anthropic|OpenAI reduces Codex Model Context Size from 372k to 272k19/07/26 · OpenAI|Introducing Cosmos 3 Edge20/07/26 · Hugging Face|Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling20/07/26 · Anthropic|At SIGGRAPH, NVIDIA Advances Graphics and Simulation With Agentic and Physical AI20/07/26 · NVIDIA|China's open-weights AI strategy is winning20/07/26|Bristol Myers Squibb Building Life Science Industry’s Most Advanced AI Factory on NVIDIA Vera Rubin20/07/26 · NVIDIA|Safety and alignment in an era of long-horizon models20/07/26 · OpenAI|Claude Fable produced a counterexample to the Jacobian Conjecture20/07/26 · Anthropic|Apply for Anthropic’s AI for Science rare disease research grants20/07/26 · Anthropic|AI advice made people less accurate but more confident – sudy19/07/26|Vincentwei1021/video-shotcraft: AI video skill for Claude Code & Codex — cinematic product videos with Remotion: 106 shot recipe cards, 161 motion previews, a production-ready template19/07/26 · Anthropic|Claude Code uses Bun written in Rust now19/07/26 · Anthropic|OpenAI reduces Codex Model Context Size from 372k to 272k19/07/26 · OpenAI|
🇨🇳 China🔬 ReasoningReleased May 2025

DeepSeek R2

DeepSeek
Open-sourceFreeSelf-host
Verdict

A solid pick for self-hosting and full data control.

Overview

DeepSeek R2 applies RL-reinforced chain-of-thought (like o3) on the V3 base. Result: reasoning performance close to o3, but open-source and 20x cheaper. The 2025 revolution for scientific AI. Slow like o3 (thinks before answering).

Skill profile

FrenchReasoningSpeedCreativitySafety
Arena Elo1370
Verified benchmarks
GPQA75.8%
SWE-Bench49.2%
MATH95.1%
MMLU89%
HumanEval92.5%

What do these scores mean?

GPQA75.8%very good

PhD-level science questions (physics, chemistry, biology), with no tool access.

SWE-Bench49.2%fair

Resolving real GitHub issues under real conditions (SWE-bench Verified).

MATH95.1%world-class

Competition-level math problems.

MMLU89%excellent

General knowledge across dozens of academic subjects.

HumanEval92.5%world-class

Python code generation from specifications.

Strengths

  • Top-tier reasoning
  • Excellent at code
  • World-class on Arena
  • Has a free tier

Limitations

  • Average speed
  • No native GDPR guarantee

Who is it for

A good fit if…
  • you need advanced reasoning or analysis
  • you want to control cost or self-host
Skip it if…
  • you handle sensitive EU data
  • you need very fast responses

Ideal use cases

  • Low-cost advanced math
  • Open-source scientific research
  • Proofs
  • Self-hosted reasoning

Access & availability

Paid API per tokenFree tier availableSelf-hostable (open weights)

Key specifications

Context
128K
Input price
$0.55 $/M
Output price
$2.19 $/M
Speed
25 tok/s
Price not auditedBenchmarks not audited

Estimate your monthly cost

Per-token API pricing
$32/ month with DeepSeek R2

For the same usage

Indicative estimate based on standard API rates (excluding caching, batch and volume discounts). Always check the official pricing before committing.

Privacy

Not GDPR-compliantHosting : CN/selfAnonymizable data
Advanced data · for experts
Architecture, modalities, detailed cost, full benchmarks
Architecture
MoE + RL
Size
685B (37B actifs MoE)
Cutoff date
Dec 2024
Inputs
text
Outputs
text, code
License
mit
Hosting
CN/self
Hallucination score
4/5
Estimated cost (API)
Typical exchange (~3k in / 1k out)
≈ $0.0038
1M in + 1M out
≈ $2.74
All benchmarks
Arena Elo1370
MMLU89.0%
GPQA75.8%
HumanEval92.5%
SWE-Bench49.2%
MATH95.1%

Similar models