Compare

Compare models

Side by side across specs, benchmarks and capability profile. Best value per row is highlighted.

Model A

Model B

Model C

Saved →

Specifications

OrganizationOpenAIDeepSeek

ReleasedDec 2025Aug 2025

WeightsClosedOpen

Parametersundisclosed671B

Active params—37B

Context400K128K

ArchitectureUndisclosed architecture (presumed dense transformer, reasoning-tuned); shipped as Instant/Thinking/Pro tiersMixture-of-Experts (671B total / 37B active) — DeepSeekMoE + Multi-head Latent Attention, hybrid think/non-think chat template

LicenseProprietaryMIT

Price in /M$1.75$0.56

Price out /M$14$1.68

VRAM @ Q4—405 GB

Elo rating2495.01903.9

Benchmarks

Aider Polyglot—76.3%

AIME100.0%93.1%

ARC-AGI-186.2%—

ARC-AGI-252.9%—

BrowseComp—30.0%

Codeforces—2091

FrontierMath (Tiers 1-3)40.3%—

GDPval70.9%—

GPQA Diamond92.4%80.1%

Humanity's Last Exam34.5%15.9%

LiveCodeBench88.9%74.8%

MMLU-Pro—84.8%

MMLU-Redux—93.7%

SWE-bench Pro55.6%—

SWE-bench Verified80.0%—

τ²-Bench Telecom98.7%—

Capability profile

Normalized against the tracked field · only categories a selected model was tested on

PREF hidden — untested by all selected models, so not charted.

GPT-5.24/62495.0

DeepSeek-V3.1 (Thinking)5/61903.9