Open-weight multimodal Mixture-of-Experts model with 2.8T total and 104B active parameters.
Human Frontier
95.8
Human-calibrated frontier signal, backed by Arena-style preference evidence and separate from raw AgMoBench benchmark composite scores.
Blended Price
$6.00/M
Input Price
$3.00/M
Output Price
$15.00/M
Speed
39 tok/s
TTFT
2.75s
Axes
Filters
Providers
v2026.06 · official-100-task · Kimi K3 (max) · harness mini-swe-agent · effort max · 1 attempt · pass@1 · harbor:snorkel-ai/senior-swe-bench-v2026.06 · provider moonshotai/kimi-k3 · record kimi-k3-mini-swe · run 45efa2b5e98d0e1b72b1d164d138bae8477c2fe17aa4a831e8999424bc24459f
{
"taskCount": 100,
"agentGroup": "mini-swe-agent:kimi-k3",
"basicPassAt1": 40.42553191489361,
"basicPassAt3": 57.446808510638306,
"datasetPackage": "snorkel-ai/senior-swe-bench-v2026.06",
"publicTaskCount": 50,
"tastefulPassAt1": 20.212765957446805,
"tastefulPassAt3": 35.1063829787234,
"agentDescription": "Moonshot Kimi K3 via mini-swe-agent harness (Portkey→OpenRouter); reasoning mandatory (max only)",
"privateTaskCount": 50,
"upstreamProvider": "kimi",
"upstreamRunCount": 50,
"averageAgentSteps": 223.80851063829786,
"averageOutputTokens": 54818.574468085106,
"averageTotalCostUsd": 7.024542418723404,
"averageOutputCostUsd": 0.8222786170212766,
"basicAllThreePassRate": 22.340425531914892,
"tastefulAllThreePassRate": 6.382978723404255
}Evidence observed 2026-07-29
v2026.06 · official-100-task · Kimi K3 (max) · harness mini-swe-agent · effort max · 1 attempt · pass@1 · harbor:snorkel-ai/senior-swe-bench-v2026.06 · provider moonshotai/kimi-k3 · record kimi-k3-mini-swe · run 132d565e2dcf4bc4526769bf487eba3743a0419460f03194909391f643ad5312
{
"taskCount": 100,
"agentGroup": "mini-swe-agent:kimi-k3",
"basicPassAt1": 35.294117647058826,
"datasetPackage": "snorkel-ai/senior-swe-bench-v2026.06",
"publicTaskCount": 50,
"tastefulPassAt1": 12.941176470588237,
"agentDescription": "Moonshot Kimi K3 via mini-swe-agent harness (Portkey→OpenRouter); reasoning mandatory (max only)",
"privateTaskCount": 50,
"upstreamProvider": "kimi",
"upstreamRunCount": 50,
"averageAgentSteps": 240.76470588235293,
"averageOutputTokens": 59167.56470588235,
"averageTotalCostUsd": 7.338732845647058,
"averageOutputCostUsd": 0.8875134705882353
}Evidence observed 2026-07-22