Status
Benchmarks
No Atlas benchmark has been run yet, so there are no results, charts or rankings here. The studies below are designed and waiting to be run.
Planned · v0.1 (protocol draft)
Coding tasks with unit tests
How often does each model produce a patch that passes the hidden unit tests of a small, licensed task set, and does a refusal-reduced variant pass more or fewer tasks than its base model?
Read the protocolPlanned · v0.1 (protocol draft)
Latency and throughput on hosted endpoints
What time to first token (TTFT) and output tokens per second does each provider endpoint deliver for a fixed request, at a stated time and region?
Read the protocolPlanned · v0.1 (protocol draft)
Instruction adherence and calibration
Do refusal-reduced variants follow formatting and factual instructions as well as their base models on a small benign set, scored by a written rubric?
Read the protocolResults: none yet.
We will publish each study with its raw outputs and scripts. Until then, the methodology explains what we will measure and how.