Shared benchmark

David DistSystemsBench

createdGolden-diff judge

29 repos: MariaDB/server, VictoriaMetrics/VictoriaMetrics, apache/flink, apache/kafka, cilium/cilium, cockroachdb/cockroach, confluentinc/librdkafka, containerd/containerd, denoland/deno, dragonflydb/dragonfly, elastic/elasticsearch, envoyproxy/envoy, facebook/rocksdb, firecracker-microvm/firecracker, grafana/mimir, istio/istio, milvus-io/milvus, mongodb/mongo, open-telemetry/opentelemetry-collector, oven-sh/bun, postgres/postgres, qdrant/qdrant, redis/redis, scylladb/scylladb, temporalio/temporal, timescale/timescaledb, valkey-io/valkey, weaviate/weaviate, yugabyte/yugabyte-db · created Aug 12, 2026 · 43 / 500 tasks

Gradingjudge_verdict

Golden-diff judge · judge_verdict

Golden-diff judge writes judge_verdict. 16 models have a rate.

Runs

Model comparisonlatest completed run · Aug 12, 14:17 UTC · judge_verdictfull comparison →

Golden-diff judge · judge_verdict


grok-4-6-mediummedium	36/43 passed	43	0
grok-4-6-highhigh	38/43 passed	43	0
grok-4-6default	39/43 passed	43	0
gemini-3.1-pro	12/43 passed	43	0
claude-sonnet-5-mediummedium	26/42 passed	43	1
claude-sonnet-5default (high)	31/42 passed	43	1
claude-sonnet-5-xhighxhigh	36/43 passed	43	0
gpt-5.6-terradefault (medium)	25/42 passed	43	1
gpt-5.6-terra-highhigh	28/43 passed	43	0
gpt-5.6-terra-xhighxhigh	32/41 passed	43	2
gpt-5.6-soldefault (medium)	33/41 passed	43	2
gpt-5.6-sol-highhigh	36/43 passed	43	0
gpt-5.6-sol-xhighxhigh	37/42 passed	43	1
claude-fable-5-mediummedium	33/43 passed	43	0
claude-fable-5default (high)	36/43 passed	43	0
claude-fable-5-xhighxhigh	38/43 passed	43	0

judge = pass rate, passed / (judged + agent failures) — a timeout, empty diff, agent error, or blown budget counts as a FAIL without a verdict (D48). grok-4.5 pinned (D22) · verdict rubric v5 (D54), the only scorer since D26. excluded = attempts outside the rate: infra failures (not the model's fault), owner-cancelled, and attempts still awaiting judge.

Embed this chart

Drop this iframe on any page. It only resolves for a sealed public benchmark, and only plots runs the owner also made public.

<iframe src="https://yourcodingbench.com/embed/benchmarks/3c6f87a4-a0e8-492d-a3ba-e594f6d2d941" title="David DistSystemsBench score vs cost" width="100%" height="480" style="border:0;border-radius:12px;overflow:hidden;background:transparent" loading="lazy"></iframe>

All runs1 run

Status	Models	Tasks	Judge	Started	Duration
completed	grok-4-6-medium, grok-4-6-high, grok-4-6, gemini-3.1-pro, claude-sonnet-5-medium, claude-sonnet-5, claude-sonnet-5-xhigh, gpt-5.6-terra, gpt-5.6-terra-high, gpt-5.6-terra-xhigh, gpt-5.6-sol, gpt-5.6-sol-high, gpt-5.6-sol-xhigh, claude-fable-5-medium, claude-fable-5, claude-fable-5-xhigh	43

Tasks

PR / commit	Repo	Title	Merged	+/−	Source
04589f9	redis/redis	Add internal connection and command mechanism (#13740)	Feb 5, 2025	+602 −28	manual
6ec38a7	oven-sh/bun