{"type":"post","id":"aa-agentperf-methodology-replays-trajectories-from-agents-resolving-cf52ce","slug":"aa-agentperf-methodology-replays-trajectories-from-agents-resolving-cf52ce","title":"AA-AgentPerf methodology replays trajectories from agents resolving GitHub issues and inserts pauses that simulate tool latency.","description":"AA-AgentPerf’s central thesis is that inference serving should be evaluated on agentic workloads rather than only on uniform prompt traffic.","retrieval_nugget":"AA-AgentPerf’s central thesis is that inference serving should be evaluated on agentic workloads rather than only on uniform prompt traffic.","published_at":"2026-10-18","updated_at":"2026-10-18","record_date":"2026-10-18","date_kind":"published_at","topics":["github"],"source_urls":["https://artificialanalysis.ai/methodology/agentperf"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:aa-agentperf-methodology-replays-trajectories-from-agents-resolving-cf52ce","status":"published","visuals":[],"editorial_provenance":{"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:ec50bcdae8aa023334e89923c65050eb432b2b7d45836d4a211aeceb0b9b1a35","reviewed_at":"2026-10-18T08:03:43.538Z","source_evidence_count":1,"verified_claim_count":2,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":2,"implication_count":1,"watch_condition_count":1,"related_record_count":0},"analysis":{"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"AA-AgentPerf’s central thesis is that inference serving should be evaluated on agentic workloads rather than only on uniform prompt traffic.","observed_facts":[{"text":"Artificial Analysis says the dataset contains coding-agent trajectories generated in the OpenCode harness using DeepSeek V3.2, GLM 4.7, and Kimi K2.5 to resolve issues in public repositories.","source_urls":["https://artificialanalysis.ai/methodology/agentperf"]},{"text":"The methodology inserts tool-call delays from under 0.1 seconds to 5 seconds, with a median of approximately one second.","source_urls":["https://artificialanalysis.ai/methodology/agentperf"]}],"mechanism":"The mechanism is trajectory replay with simulated tool latency: after each tool call, the benchmark inserts a delay sampled from a distribution partitioned by tool type.","why_now":"Why now: as inference providers serve more multi-step coding-agent traffic, benchmark rankings based on isolated requests may not capture the interaction between concurrency, context reuse, scheduler behavior, and time spent waiting on tools.","implications":["The practical implication is that teams choosing serving configurations should ask whether their evaluation workload resembles their agents’ actual trajectories, not just their token mix."],"evidence_boundary":"Evidence boundary: the cited page documents the benchmark’s design and metrics; it does not, by itself, establish that any system is faster, cheaper, or more suitable than another.","watch_conditions":["Watch whether published leaderboard entries disclose comparable configurations and whether results remain stable across model and software-stack updates."],"related_records":[],"new_branch_reason":"This begins a distinct analytical record because no prior New Runtime record was supplied for this specific methodology."},"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/aa-agentperf-methodology-replays-trajectories-from-agents-resolving-cf52ce/","markdown":"https://newruntime.com/posts/aa-agentperf-methodology-replays-trajectories-from-agents-resolving-cf52ce.md","json":"https://newruntime.com/posts/aa-agentperf-methodology-replays-trajectories-from-agents-resolving-cf52ce.json"},"related_patterns":["agent-economics-moves-to-completed-work"],"next_reads":[{"type":"topic","path":"/topics/github/","url":"https://newruntime.com/topics/github/","title":"Github","reason":"Explore the github topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/github.md","json_url":"https://newruntime.com/topics/github.json"},{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/source-ledger.json","url":"https://newruntime.com/source-ledger.json","title":"Source Ledger","reason":"Inspect the public source evidence graph.","media_type":"application/json"}]}