whatmodel.sh / Nemotron 3.5 Bolt 74B A9B
Nemotron 3.5 Bolt 74B A9B
↗ share
+ contribute
Nemotron 3.5 Bolt 74B A9B is an open-weights language model from NVIDIA, released October 2026. Its published size is 74B parameters (9B active per token) and a 262k-token context window. Licence: open-mdw-1.1.
Catalogue identity notes
Precision is retained in each measurement: lab rows BF16; Super VL Arena rows NVFP4. Quantized deployment is not a separate checkpoint ranking.
Model information sources
Score within the LMArena (text) comparison set only.
$ how it compares
The LLMs ranked either side of Nemotron 3.5 Bolt 74B A9B, out of 882 with a score, then more from NVIDIA.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 105 |  Ring-2.6-1T | 68±2 | comparison only | ||
| 106 | Nemotron 3.5 Super VL 120B A12B | 68±1 | local | ||
| 107 | Nemotron 3.5 Bolt 74B A9B | 68±2 | local | ||
| 108 |  A.X K2 (688B-A33B) | 68±1 | local | ||
| 109 |  MiMo-V2.5 (310B-A15B) | 68±1 | api $0.21/M +local | ||
| more from NVIDIA | |||||
| 102 | Nemotron 3 Ultra 550B-A55B | 69±1 | api $1.35/M +local | ||
| 168 | Nemotron Cascade 2 30B-A3B | 65±1 | local | ||
| 179 | Nemotron 3 Super 120B-A12B | 64±1 | api $0.28/M +local | ||
$ what it's good at
agent tasks▁▁▃▆█▅▅▅▅▄▂▁ #83 of 310 · 39% of its score · 3 tests
Beats GPT-5, Grok 4.20 (0309 reasoning), GPT-5.1 here, though they rank higher overall. Trails MiMo-V2.5 (310B-A15B) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Terminal-Bench 2.1 lab | 53.9% ✎ | ▁·▁·▁▁▄▃▅█▅█ beats 19/61 | 20% |
| BrowseComp lab | 70.0% ✎ | ▂·▃▁▂▂▆▄█▂▂▄ beats 33/44 | 13% |
| τ³-bench Banking lab | 17.0% ✎ | ▂▂·▅█▂▃▃▃▆▂▅ beats 16/24 | 6.0% |
codingnot ranked on this task yet · 1 test
| test | result | vs the others in its run | counts |
|---|---|---|---|
| SciCode (level unspecified) lab | 44.5% ✎ | ▁····▃▄▄▅█▇▃ beats 32/37 | — |
web search█··········█ #2 of 2 * · 13% of its score · 1 test
| test | result | vs the others in its run | counts |
|---|---|---|---|
| BrowseComp lab | 70.0% ✎ | ▂·▃▁▂▂▆▄█▂▂▄ beats 33/44 | 13% |
other tests3 not in a task · 59% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GPQA Diamond lab | 88.8% ✎ | ▁▁▁▁▂▄▅▅█▇█▄ beats 258/315 | 34% |
| Humanity's Last Exam lab | 28.7% ✎ | ▁▁▁▄▅▆▆██▆▁▃ beats 69/108 | 25% |
| MCP-Atlas v1 public500 lab | 55.6% ✎ | █·····█····█ beats 0/2 | — |
$ where to run it
Other ways to run it
- Run it as published · needs ~148 GB, NVIDIA only · from huggingface.co/nvidia (official)
wget https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16/resolve/main/ultra_v3_reasoning_parser.py export VLLM_USE_V2_MODEL_RUNNER=1 vllm serve nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16 \ --trust-remote-code \ --load-format auto \ --tensor-parallel-size 1 \ --max-model-len 262144 \ --max-num-seqs 1 \ --enable-chunked-prefill \ --attention-backend FLASHINFER \ --mamba-backend flashinfer \ --mamba-cache-mode align \ --seed 42 \ --allowed-local-media-path / \ --limit-mm-per-prompt '{"video":1,"image":1}' \ --video-pruning-rate 0.5 \ --media-io-kwargs '{"video":{"fps":2,"num_frames":256}}' \ --enable-auto-tool-choice \ --tool-call-parser qwen3_coder \ --reasoning-parser-plugin ultra_v3_reasoning_parser.py \ --reasoning-parser ultra_v3
Official weights and documentation: huggingface.co/nvidia ↗
$ how the score was worked out · NVIDIA vendor-reported benchm… 98%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 NVIDIA vendor-reported benchmark scores for its own Nemotron models (Hugging Face model cards) · lab-reported | 5 | 98% |
| prior · pull toward the middle | 2% | |
| weighted average | = 68.4 |
not counted (2)
| NVIDIA vendor-reported benchmark scores for its own Nemotron models (Hugging Face model cards) | SciCode (level unspecified) [correct] | 44.5% | fewer than three models have this test yet, or this source ran it for this model alone |
| NVIDIA vendor-reported benchmark scores for its own Nemotron models (Hugging Face model cards) | MCP-Atlas v1 public500 [correct] | 55.6% | fewer than three models have this test yet, or this source ran it for this model alone |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json