LetTokenBurnA benchmark with 406 tasks and 4,836 criteria from 23 simulated organizations across 8 domains, measuring agent performance on professional work.
$ npm i -g lettokenburn-cli$ ltb bench info officetown-bench
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| GPT-6 Astra | 74.4 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| Grok-4.7 | 58.4 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| Claude Opus 5.5 | 50.7 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| GPT-5.6-terra | 39.4 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| GPT-5.6 Sol | 33.5 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| GPT-5.6 Luna | 32.5 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| GPT-5.4 | 30.5 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| GPT-5.2-chat | 21.2 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| DeepSeek V4 Pro | 18.5 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| DeepSeek V4 Flash | 13.5 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| Grok-4.3 | 8.1 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| Kimi K2.5 | 4.9 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| Mistral Large 3 | 1.2 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Constructs executable resources from generative organizational simulations, incorporating domain-expert feedback into task and rubric refinement.
Evaluates frontier models using rubric scores, revealing substantial headroom and showing improvements via supervised fine-tuning.