prague-bench

← Failures

Kimi K3.1 Pro

Did not finish · 5-hour plan limitMoonshot AIOpenCode · OpenCode GoEffort: high13 Nov 2026spec 666c07f57d6e

No video

Key numbers

Panel score
–not scored
Cost
$2.84API list price (paid by a subscription)
Wall time
23 min
Model time
22 min96% of wall time
tok/s
58.8
Tokens
8.8MOutput: 77k
Cache hit
98%
API calls
80
Tool calls
90
Renderer
Intel UHD 630 (GPU)
H1–H6
No video

How the run ended

Did not finish: the OpenCode Go subscription hit its 5-hour usage limit in the middle of a tool call. A full render (1740 frames) had already finished, but the agent was still iterating and had not encoded the MP4. Worth a re-run.

Time and speed

Wall time includes rendering and encoding. Model time is the sum of all model calls.

  • Model generating
  • Tools & waiting
  1. Wall time23 min96% model
Time and speed
Wall time23 min
Model time22 min
Tools & waiting1.0 min
Model share of wall time96%
tok/s58.8
Model calls80

Tokens and cost

Tokens summed over every model response. Tokens multiplied by the provider’s price.

Tokens and cost
Uncached input8,695
Cache read8,561,664
Cache write167,167
Output77,461of which 30,657 reasoning
Total8,814,987
Cache-hit ratio98.0%
Cost$2.84API list price (paid by a subscription)

Tools used

Tool calls by name, as the harness logged them. Names differ between harnesses.

  1. shell28
  2. edit25
  3. read18
  4. write17
  5. grep2

Rendering

Rendering
RendererIntel UHD 630 (GPU)
Final render2.7 min
Per frame–

Hard requirements H1–H6

No video was delivered, so there was nothing to check.

From the agent’s report

The agent did not write a REPORT.md.

Contact sheet and route map