Code Generation on LiveCodeBench v6 (2025-02 to 2025-05)

74.1Accuracy

Qwen3-235B-A22B-Thinking-2507

Updated 21d ago

Evaluation Results

Method
Qwen3-235B-A22B-Thinking-2507 2026.05	74.1	-	-	-
DuST 2026.05	72.6	100	82.3	52.5
Gemini-2.5 Pro 2026.05	72.5	-	-	-
PaCoRe hybrid compaction variant 2026.05	71.1	-	-	-
DuST 2026.05	69.4	98.7	78.5	47.1
ZAYA1-8B + Markovian RSA (40K/4K) 2026.05	69.2	-	-	-
DeepSeek-R1-0528 2026.05	68.7	-	-	-
Qwen3-30B-Thinking 2026.05	68.7	100	73.7	47.5
Qwen3-Next-80B-A3B-Think 2026.05	67.8	-	-	-
OpenReasoning-Nemotron-32B 2026.05	67.4	98.4	73.1	48
Intellect-3 2026.05	66.8	-	-	-
GPT-OSS-20B 2026.05	65.2	98.2	75.2	43.5
ZAYA1-8B (single rollout) 2026.05	65	-	-	-
TinyR1-32B 2026.05	64.9	89.5	74.4	46.3
OpenCodeReasoning-1.1-32B 2026.05	64.9	97.6	71.2	44.3
ZAYA1-8B 2026.05	64.8	-	-	-
Nemotron-3-Nano-30B-A3B 2026.05	64.6	-	-	-
OpenReasoning-Nemotron-14B 2026.05	63	97.6	72.4	39.3
OpenCodeReasoning-1.1-14B 2026.05	60.3	96.8	65.4	38.5
DuST 2026.05	59.4	98.4	68	34
OpenReasoning-Nemotron-7B 2026.05	59	99.2	66	34
OLMo-3.1-32B-Think 2026.05	58.3	-	-	-
Mistral-Small-4-119B 2026.05	57.9	-	-	-
Klear-Reasoner-8B 2026.05	57.6	98.4	64.7	32.4
Qwen3-4B-Thinking 2026.05	55	97.6	60.3	29.9
DeepSeek-R1-0528-Qwen3-8B 2026.05	53.8	97.6	56.4	29.9
DuST 2026.05	47.1	88.2	49.6	24.6
OpenCodeReasoning-1.1-7B 2026.05	45	94.4	55.8	13.1
Qwen3-30B-Instruct 2026.05	43.1	82.9	49.2	19
Skywork-OR1-7B 2026.05	40.5	89.5	41	15.2
Arcee-Trinity-Mini 2026.05	33.3	-	-	-