Measured, not claimed

Against the others, on the same tasks

Recorded task outcomes from 2026-09-17. This comparison is exploratory: matching models, resource budgets and independent task coverage have not been verified. Numerical leaders describe these recorded results; they do not establish a competitor win percentage.

The improvement race

See the lead. See the next challenge.

Explore the latest published measurements. Select a category, then focus or tap a bar for its evidence. Highlighted segments show a measured advantage.

Loading the published comparison…

Read the recorded benchmark tables.

Published results refresh here every minute. A new score requires a completed benchmark and publication; animation does not indicate a new measurement.

Comparison on corpus 1.0.0 (2026-09-17)

Exploratory recorded results. Metric leaders describe the available numbers, not confirmed product superiority. Matching model, resource budgets, independent tasks and uncertainty have not been verified. Missing metrics are not wins.

metric	better	priestai	claude-code	ollama-alone	openclaw-local	openclaw-cloud	leader
correctness	higher	0.867	1.0	0.0	0.533	0.6	Indicative only: unequal trials (priestai 3, claude-code 1, ollama-alone 3, openclaw-local 3, openclaw-cloud 3). No metric leader.
completionRate	higher	1.0	1.0	1.0	1.0	1.0	Indicative only: unequal trials (priestai 3, claude-code 1, ollama-alone 3, openclaw-local 3, openclaw-cloud 3). No metric leader.
secondsPerVerifiedTask	lower	161.61	14.553	no verified work	130.066	83.932	Indicative only: unequal trials (priestai 3, claude-code 1, ollama-alone 3, openclaw-local 3, openclaw-cloud 3). No metric leader.
endToEndSecondsMedian	lower	115.843	14.437	no verified work	62.812	47.187	Indicative only: unequal trials (priestai 3, claude-code 1, ollama-alone 3, openclaw-local 3, openclaw-cloud 3). No metric leader.
timeToFirstResultSeconds	lower	90.656	not measured	no verified work	not measured	not measured	Indicative only: unequal trials (priestai 3, claude-code 1, ollama-alone 3, openclaw-local 3, openclaw-cloud 3). No metric leader.
strayFilesPerTask	lower	0.0	1.6	no verified work	0.0	0.0	Indicative only: unequal trials (priestai 3, claude-code 1, ollama-alone 3, openclaw-local 3, openclaw-cloud 3). No metric leader.
invalidToolCallRate	lower	0.226	not measured	no verified work	not measured	not measured	Indicative only: unequal trials (priestai 3, claude-code 1, ollama-alone 3, openclaw-local 3, openclaw-cloud 3). No metric leader.
regressionRate	lower	0.133	0.0	no verified work	0.067	0.0	Indicative only: unequal trials (priestai 3, claude-code 1, ollama-alone 3, openclaw-local 3, openclaw-cloud 3). No metric leader.

Per task (passed/trials)

task	priestai	claude-code	ollama-alone	openclaw-local	openclaw-cloud
fix-typo	3/3	1/1	0/3	1/3	3/3
write-test	2/3	1/1	0/3	1/3	3/3
read-only-answer	3/3	1/1	0/3	0/3	0/3
rename-across-files	2/3	1/1	0/3	3/3	0/3
make-the-test-pass	3/3	1/1	0/3	3/3	3/3

Strengths and weaknesses

Indicative only: unequal trials (priestai 3, claude-code 1, ollama-alone 3, openclaw-local 3, openclaw-cloud 3). No strength or weakness is named.

Disclosure

priestai: route kiwi_kiwi/mythos-9b-merged_v2-abliterated:q4k; trials 3; hardware laptop; served {'model': 'kiwi_kiwi/mythos-9b-merged_v2-abliterated:q4k', 'host': 'local', 'asked': '', 'ctx': 65536, 'ctx_measured': True}; this product; warm-up not recorded (result predates the warm-up phase; any model load is inside its task seconds); configuration {"agentCommand": "", "agentEnvKeys": [], "externalAgent": false, "hardware": {"cpuCount": 32, "gpuMemoryGb": null, "memoryGb": 15.6, "platform": "win32"}, "label": "priestai", "mode": "edits", "productVersion": "0.1.0"}
claude-code: route external:claude; trials 1 (indicative: under three); hardware laptop; served not reported; external agent through a subprocess (tool calls not observable); warm-up not recorded (result predates the warm-up phase; any model load is inside its task seconds); configuration {"agentCommand": "claude", "agentEnvKeys": [], "externalAgent": true, "hardware": {"cpuCount": 32, "gpuMemoryGb": null, "memoryGb": 15.6, "platform": "win32"}, "label": "claude-code", "mode": "edits", "productVersion": "0.1.0"}
ollama-alone: route external:ollama; trials 3; hardware laptop; served not reported; external agent through a subprocess (tool calls not observable); warm-up not recorded (result predates the warm-up phase; any model load is inside its task seconds); configuration {"agentCommand": "ollama", "agentEnvKeys": [], "externalAgent": true, "hardware": {"cpuCount": 32, "gpuMemoryGb": null, "memoryGb": 15.6, "platform": "win32"}, "label": "ollama-alone", "mode": "edits", "productVersion": "0.1.0"}
openclaw-local: route external:openclaw; trials 3; hardware laptop; served not reported; external agent through a subprocess (tool calls not observable); warm-up not recorded (result predates the warm-up phase; any model load is inside its task seconds); configuration {"agentCommand": "openclaw", "agentEnvKeys": ["OLLAMA_API_KEY"], "externalAgent": true, "hardware": {"cpuCount": 32, "gpuMemoryGb": null, "memoryGb": 15.6, "platform": "win32"}, "label": "openclaw-local", "mode": "edits", "productVersion": "0.1.0"}
openclaw-cloud: route external:openclaw; trials 3; hardware laptop; served not reported; external agent through a subprocess (tool calls not observable); warm-up not recorded (result predates the warm-up phase; any model load is inside its task seconds); configuration {"agentCommand": "openclaw", "agentEnvKeys": [], "externalAgent": true, "hardware": {"cpuCount": 32, "gpuMemoryGb": null, "memoryGb": 15.6, "platform": "win32"}, "label": "openclaw-cloud", "mode": "edits", "productVersion": "0.1.0"}

The harness that produced this page ships inside every download: python priest_cli.py eval run runs the corpus on this machine, --agent-command runs any other agent on it, and eval compare writes a page like this one.