Compare AI models
| Specifications | ||
ParametersThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. | — | 671B |
Context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider. | 200k | 128k |
| Benchmarks | ||
SWE-Bench VerifiedCoding — Real coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. | 40.6% | 49.2% |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 41.6% | 71.5% |
MMLUGeneral knowledge — A 57-subject multiple-choice exam — history, law, medicine, maths — that was the standard measure of how much a model knows from 2020 until roughly 2024, when frontier scores crowded into the high 80s and labs moved on to harder tests. The scores here were published years apart under different testing setups, so read them as a historical record rather than a like-for-like ranking. Higher is better. | 80.9% | 90.8% |
BullshitBench v2Nonsense detection — Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better. | 50% | — |
LiveCodeBenchCompetitive coding — Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better. | — | 65.9% |
HumanEvalFunction synthesis — 164 small Python problems: the AI is given a function's description and has to write the working function. This was the coding benchmark of the GPT-3.5 and GPT-4 era, before the field moved to fixing real bugs in real repositories. Higher is better. | 88.1% | — |
MATH-500Reasoning — accuracy on the 500-problem MATH subset | — | 97.3% |
DROPReasoning — three-shot token F1 on questions requiring reasoning over paragraphs | — | 92.2 |
AIME 2024Reasoning — single-answer accuracy on the 2024 invitational mathematics exam without tools | — | 79.8% |
SimpleQAKnowledge — accuracy on short, fact-seeking questions in the original SimpleQA set | — | 30.1% |
MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark | — | 84% |
MGSMMultilingual math — solving grade-school word problems translated into multiple languages | 85.6% | — |
| Overview | ||
| Company | Anthropic | DeepSeek |
| Release date | Oct 22 2024 | Jan 20 2025 |
| Access | Closed | Open Weight |
| Model details | View model | View model |
Other comparisons
Claude 3.5 HaikuvsGPT-6.2 AstraDeepSeek R1vsGPT-6.2 AstraClaude 3.5 HaikuvsGemini 4 HydrogenDeepSeek R1vsGemini 4 HydrogenClaude 3.5 HaikuvsMuse EmberDeepSeek R1vsMuse EmberClaude 3.5 HaikuvsGrok 4.7DeepSeek R1vsGrok 4.7Claude 3.5 HaikuvsMistral Large 4DeepSeek R1vsMistral Large 4Claude 3.5 HaikuvsKimi K3.1 ProDeepSeek R1vsKimi K3.1 ProFrequently asked questions
DeepSeek R1 leads Claude 3.5 Haiku on 3 of the 3 benchmarks they both report (SWE-Bench Verified, GPQA Diamond, MMLU). Claude 3.5 Haiku shipped 90 days before DeepSeek R1, so benchmark comparisons should account for the intervening progress.
Context windows are 200k (Claude 3.5 Haiku) vs 128k (DeepSeek R1). Claude 3.5 Haiku is closed, while DeepSeek R1 is open weight.
On SWE-Bench Verified, DeepSeek R1 leads at 49.2% vs Claude 3.5 Haiku at 40.6%. On GPQA Diamond, DeepSeek R1 leads at 71.5% vs Claude 3.5 Haiku at 41.6%. On MMLU, DeepSeek R1 leads at 90.8% vs Claude 3.5 Haiku at 80.9%.