chi-bench: Actava · openai-agents · moonshotai/kimi-k3 - #19
Merged
haolin-chen-actava merged 1 commit intoAug 12, 2026
Merged
Conversation
Filed 2026-08-12 from the original 2026-07-22/24 run; aggregation unmodified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Submission validation —
|
haolin-chen-actava
deleted the
sub/chi-bench/2026-07-24-kimi-k3-openai-agents
branch
August 12, 2026 20:20
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Kimi K3 on the OpenAI Agents SDK harness, from the 2026-07-24 frontier release wave.
Full-run cost $95.36. Matches GPT-5.6 Sol on pass@1 at $59.16 less.
Why this is filed late. The run completed 2026-07-22 and its results were published in the 4/4 results post on 2026-07-29, but the leaderboard submission was never filed. This packet is the original, unmodified aggregation — the numbers are identical to what was published.
Cost basis. All 75 trials carry price-table-estimated cost rather than provider-reported cost (
cost_reported_trials = 0), and 4 pa_um trials terminated with agent exceptions and report no usage. Leaderboard cost is therefore a lower bound on billed execution cost.Trajectory repair. Those same 4 trials completed verification but emitted no
trace.jsonl. Error-only ATIF-v1.2 trajectories were generated with the officialopenai_agents_harness._build_atif_trajectory(records=[], ...)builder from each preservedagent/run_result.json; original logs were not modified. This is the same treatment applied to the Inkling submission.chi_bench_git_shapins2395e9e, now onmainvia actava-ai/chi-bench#42.Validator: 0 errors, 0 warnings.
🤖 Generated with Claude Code