Run -Qt9WkAJt76K
Eval: pepper-grinder 2.0 kstrat2001/pepper-grinder@23d51a2
This is the eval's reference run, recorded by a maintainer on approval.
Measurements
- Outcome
- Passed
See what it did: sign in or create a free account.
- Status
- Submitted
- Active time
- 3m 15s
- Tokens per second
- 115.4
- Peak memory
- not reported
- darkmux version
- 5.3.0
- Shared
- 2026-10-22
Why it passed or failed
- Verify said
- With a free account
- Tests before
- With a free account
- Tests after
- With a free account
- Passing
- With a free account
- Tests added
- With a free account
- Failing
- With a free account
- Changed the code
- With a free account
- Test command exit
- With a free account
The run
- Wall clock
- With a free account
- Thermal rests
- With a free account
- Turns
- With a free account
- Tool calls
- With a free account
- Tool calls that failed
- With a free account
- Tokens generated
- With a free account
- Of them, reasoning
- With a free account
The machine while busy
- Package power
- With a free account
- CPU power
- With a free account
- GPU power
- With a free account
- GPU busy
- With a free account
- Energy per 1,000 tokens
- With a free account
- Memory pressure, highest
- With a free account
- Thermal states
- With a free account
- CPU speed limit, lowest
- With a free account
- Throttled samples
- With a free account
A free account adds
- Why each run passed or failed: what verify said, and the tests before and after
- Power, thermals and loads while the model worked
- Turns, tool calls and tokens
- The range around each median, from the fastest run to the slowest
- The trajectory, every step the model took
- Submitting your own runs (by invitation while the first results go in)
One darklyenergized.com account, free. Sign in and you come back to this page.
What it ran on
- Hardware class
- Apple M5 Max, 128 GB
- Chip
- Apple M5 Max
- Memory
- 128 GB
- Accelerator
- Metal
- Operating system
- macOS 26.6.2
- Engine
- rec
- Engine backend
- MLX
- Engine version
- 0.1.0
- Model
- flint
- Quantization
- int4
- Weights hash
- fad30edd46369f56bfe34a80a6d8ab0f416b7f02af2d10dab6f5d75166d0d16c
- Context length
- 262144
- KV cache type
- not reported
Trajectory
No trajectory was shared with this run.