These scores were compiled from the linked sources. These published scores were not independently tested here. Different prompts, precision, thinking budgets and tools can change results.
OSWorld · Holo harness
80.8 / 100
Computer-use task success in H Company's harness. Mean over two to four publisher runs; harness and benchmark release can affect results.
Benchmark efficiency: 3.794 published score points per GB of selected weights. This ratio is not an inference-speed measure; check the test settings to see whether the score covers the selected quantization.
Test settings: Publisher evaluation in the Holo4 harness; OSWorld task success, mean over two to four runs. The announcement does not name a pinned OSWorld release in this result row.
Original model result, not an independent test of the selected Q4 package. The ranked 21,297,870,912-byte language weight excludes the required 899,282,976-byte F16 vision projector. Local multimodal use also needs the H agent harness.
Average partial score on OSWorld 2.0 in H Company's harness. Single publisher run; distinct from task success rate and other harnesses.
Benchmark efficiency: 1.451 published score points per GB of selected weights. This ratio is not an inference-speed measure; check the test settings to see whether the score covers the selected quantization.
Test settings: Publisher single run in the Holo4 harness; OSWorld 2.0 average partial score. The associated task-success rate was 12.3%, which is not substituted here.
Original model result, not an independent test of the selected Q4 package. The required 899,282,976-byte F16 vision projector is excluded from ranked language-weight bytes. Harness, benchmark release and generation budget affect comparisons.
H Company's 120-task held-out split from AutomationBench v1.0.6. Kept separate from the full 600-task public set because 480 public tasks overlapped the training-data split.
Benchmark efficiency: 1.488 published score points per GB of selected weights. This ratio is not an inference-speed measure; check the test settings to see whether the score covers the selected quantization.
Test settings: Publisher score on the 120 held-out tasks from AutomationBench v1.0.6. The other 480 public tasks fell in the split used to collect training data and are excluded from this view.
Custom held-out split, not the official private leaderboard. Original model result; selected Q4 package was not independently retested. Vision use also needs the separate 899,282,976-byte projector.
Context and version evidence: The repository and file name identify the selected model. Revision hashes are shown above when recorded. Tested context length, full hardware configuration and run date are not consistently recorded in this snapshot. Only settings explicitly stated in each benchmark record are known.
The 21.298 GB figure covers model weights in this file. It excludes the context cache, runtime, operating system and any separate multimodal files. It is not a measured RAM requirement or a token-generation speed.
These complete configurations pass the initial weight-size check for the selected 21.30 GB file. Allow extra memory for context and software. Check the model’s runtime notes and your inference software before buying.
Listing evidence: 2026-09-25. Amazon.com base price visible in an owner-supplied screenshot reviewed 25 Sep 2026; capture time not supplied. Dated reference, not a live quote. Excludes card signup offers, coupons, tax and shipping.
Listing evidence: 2026-10-02. Apple US launch price for the base M5 Max Mac Studio, published 22 September 2026: USD 2,499. Dated manufacturer reference, not an Amazon offer; tax and shipping may differ.
Listing evidence: 2026-09-24. GMKtec, EVO-X2 / 64GB / 1TB, recorded 22 Sep 2026; exact ASIN capacity identified 24 Sep. This is a dated reference, not a current Amazon price; tax, shipping and bundle contents may differ.
Memory matches are estimates, not tested compatibility or speed. As an Amazon Associate I earn from qualifying purchases.
Confirm model architecture and quantization support in your runner. The selected weights use 21.298 GB; this excludes context cache, runtime buffers and other programs.
These are the first listed tiers that pass our reserve policy, not measured minimum requirements. System RAM and VRAM remain separate. Use the memory tool to change the reserve for your context and runner.