Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
1-bit Bonsai 4B
~44
2/8 categoriesDeepSeek-R1
45
Winner · 2/8 categories1-bit Bonsai 4B· DeepSeek-R1
Pick DeepSeek-R1 if you want the stronger benchmark profile. 1-bit Bonsai 4B only becomes the better choice if mathematics is the priority or you want the cheaper token bill.
DeepSeek-R1 finishes one point ahead overall, 45 to 44. That is enough to call, but not enough to treat as a blowout. This matchup comes down to a few meaningful edges rather than one model dominating the board.
DeepSeek-R1's sharpest advantage is in knowledge, where it averages 47 against 28.7. The single biggest benchmark swing on the page is GPQA, 28.7% to 71.5%. 1-bit Bonsai 4B does hit back in mathematics, so the answer changes if that is the part of the workload you care about most.
DeepSeek-R1 is also the more expensive model on tokens at $0.55 input / $2.19 output per 1M tokens, versus $0.00 input / $0.00 output per 1M tokens for 1-bit Bonsai 4B. That is roughly Infinityx on output cost alone. DeepSeek-R1 is the reasoning model in the pair, while 1-bit Bonsai 4B is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. DeepSeek-R1 gives you the larger context window at 128K, compared with 32K for 1-bit Bonsai 4B.
BenchLM keeps the benchmark table and the operator tradeoffs on the same page so a better score does not hide a materially slower, pricier, or smaller-context model.
Runtime metrics show N/A when BenchLM does not have a sourced snapshot for that exact model. The scoring rules and freshness policy are documented on the methodology page.
| Benchmark | 1-bit Bonsai 4B | DeepSeek-R1 |
|---|---|---|
| Agentic | ||
| Terminal-Bench 2.0 | — | 42% |
| BrowseComp | — | 49% |
| OSWorld-Verified | — | 44% |
| Coding | ||
| HumanEval | — | 92% |
| SWE-bench Verified | — | 49.2% |
| LiveCodeBench | — | 19% |
| SWE-bench Pro | — | 25% |
| Multimodal & Grounded | ||
| MMMU-Pro | — | 43% |
| OfficeQA Pro | — | 53% |
| Reasoning1-bit Bonsai 4B wins | ||
| MuSR | 41.4% | 40% |
| BBH | — | 66% |
| LongBench v2 | — | 58% |
| MRCRv2 | — | 57% |
| ARC-AGI-2 | — | 1.3% |
| KnowledgeDeepSeek-R1 wins | ||
| GPQA | 28.7% | 71.5% |
| MMLU | — | 90.8% |
| SuperGPQA | — | 41% |
| MMLU-Pro | — | 84% |
| HLE | — | 14% |
| FrontierScience | — | 44% |
| SimpleQA | — | 30.1% |
| Instruction FollowingDeepSeek-R1 wins | ||
| IFEval | 69.6% | 83.3% |
| Multilingual | ||
| MGSM | — | 61% |
| MMLU-ProX | — | 60% |
| Mathematics1-bit Bonsai 4B wins | ||
| MATH-500 | 65.8% | 97.3% |
| AIME 2023 | — | 44% |
| AIME 2024 | — | 79.8% |
| AIME 2025 | — | 45% |
| HMMT Feb 2023 | — | 40% |
| HMMT Feb 2024 | — | 42% |
| HMMT Feb 2025 | — | 41% |
| BRUMO 2025 | — | 43% |
DeepSeek-R1 is ahead overall, 45 to 44. The biggest single separator in this matchup is GPQA, where the scores are 28.7% and 71.5%.
DeepSeek-R1 has the edge for knowledge tasks in this comparison, averaging 47 versus 28.7. Inside this category, GPQA is the benchmark that creates the most daylight between them.
1-bit Bonsai 4B has the edge for math in this comparison, averaging 65.8 versus 57.4. Inside this category, MATH-500 is the benchmark that creates the most daylight between them.
1-bit Bonsai 4B has the edge for reasoning in this comparison, averaging 41.4 versus 40. Inside this category, MuSR is the benchmark that creates the most daylight between them.
DeepSeek-R1 has the edge for instruction following in this comparison, averaging 83.3 versus 69.6. Inside this category, IFEval is the benchmark that creates the most daylight between them.
Get notified when new models drop, benchmark scores change, or the leaderboard shifts. One email per week.
Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.