Lab vs Lab
NVIDIA vs Thinking Machines Lab.
The club-level face-off: Club Scores, best models, squad depth, division titles and the price floor — every number computed from the model database.
Club face-off | NVIDIA | Thinking Machines | Edge |
|---|
| Club Score | 45.7 | 43.1 | NVIDIA |
| Best model | Nemotron 3 Ultra 550B-A55B | Inkling | NVIDIA |
| Top-3 average | 28.6 | 43.1 | Thinking Machines |
| Squad size | 3 | 2 | NVIDIA |
| Cheapest capable ($/1M blended) | — | — | — |
| Open-weight share % | 100 | 100 | Draw |
| Division titles | 0 | 0 | Draw |
Price floor — cheapest capable model on three workloads
| Workload | NVIDIA | Thinking Machines |
|---|
| Chat assistant (1M in / 250K out) | — | — |
| Document analysis (5M in / 500K out) | — | — |
| Code generation (2M in / 1000K out) | — | — |
Best vs best
League fixture: The two clubs' #1 models, head-to-head. In Season Zero, fixture results are decided by the data — the category winners above. When voted fixtures launch, results blend 60% benchmark / 40% crowd per the published rulebook.
Verdict by category
| Category | Nemotron 3 Ultra 550B-A55B | Inkling | Winner |
|---|
| League Score | 57.1 | 43.1 | Nemotron 3 Ultra 550B-A55B |
| Coding (SWE-bench %) | 71.9 | — | — |
| Reasoning (GPQA %) | 87.0 | — | — |
| Context (tokens) | 1000000 | 262144 | Nemotron 3 Ultra 550B-A55B |
| Speed (tok/s) | — | — | — |
Tale of the tape
| Spec | Nemotron 3 Ultra 550B-A55B | Inkling |
|---|
| Maker | NVIDIA | Thinking Machines Lab |
| Released | Jun 4, 2026 | Jul 15, 2026 |
| League Score | 57 | 43 |
| Context | 1,000,000 tokens | 262,144 tokens |
| Max output | — | — |
| $/1M input | — | — |
| $/1M output | — | — |
| Weights | Open | Open |
| Speed | — | — |
Head-to-head benchmarks (shared only)
| Benchmark | Nemotron 3 Ultra 550B-A55B | Inkling |
|---|
| No shared benchmark measurements yet — shown as not measured, never interpolated. |
Cost at three workloads
| Workload | Nemotron 3 Ultra 550B-A55B | Inkling |
|---|
| Light — 1M in / 250K out | — | — |
| Standard — 10M in / 2.5M out | — | — |
| Long-context — 50M in / 5M out | — | — |
The pick
Pick Nemotron 3 Ultra 550B-A55B if you want the higher overall League Score; you need the bigger context window; coding is the priority.
All figures from the model database — sources and dates on each model page. Cost your own workload or compare a different pair.
NVIDIA profile Thinking Machines profile Compare other labs
