live
weekly refresh
basedagi.org
benchmark evidence

LiveBench Coding

LiveBench coding benchmark.

winner on LiveBench Coding
direct benchmark result, not a broad vertical composite | source row dated 2026-05-20
scored on 2026-05-20
latest mapped results | top 20
#ModelScoreEvidenceTested
1DeepSeek: DeepSeek V3.2
Deepseek
71.1
model-only
independent_benchmark
2026-05-20
2DeepSeek: R1
Deepseek
70.3
model-only
independent_benchmark
2026-05-20
3OpenAI: o3 Mini
Openai
69.5
model-only
independent_benchmark
2026-05-20
4Google: Gemini 2.5 Pro
Google
63.3
model-only
independent_benchmark
2026-05-20
5Qwen2.5 Coder 32B Instruct
Qwen
57.0
model-only
independent_benchmark
2026-05-20
6Google: Gemini 2.0 Flash
Google
54.7
model-only
independent_benchmark
2026-05-20
7OpenAI: o1
Openai
52.3
model-only
independent_benchmark
2026-05-20
8Meta: Llama 4 Maverick
Meta Llama
51.1
model-only
independent_benchmark
2026-05-20
9Mistral Large
Mistralai
46.9
model-only
independent_benchmark
2026-05-20
10OpenAI: GPT-5
Openai
46.1
model-only
independent_benchmark
2026-05-20
11DeepSeek: DeepSeek V4 Pro
Deepseek
44.4
model-only
independent_benchmark
2026-05-08
12Google: Gemini 2.0 Flash Lite
Google
43.8
model-only
independent_benchmark
2026-05-20
13Google: Gemma 3 27B
Google
39.8
model-only
independent_benchmark
2026-05-20
14Amazon: Nova Pro 1.0
Amazon
39.1
model-only
independent_benchmark
2026-05-20
15Mistral: Mistral Small 4
Mistralai
36.7
model-only
independent_benchmark
2026-05-20
16Meta: Llama 3.3 70B Instruct
Meta Llama
20.0
model-only
independent_benchmark
2026-05-20
what this result means

LiveBench coding benchmark.

This benchmark contributes direct public evidence. Read its scope before generalizing the result.

A win here is a win on LiveBench Coding. Broad task pages require independent corroboration before naming a general winner.

source record
category: coding
metric: accuracy
matched models: 16
latest source date: 2026-05-20
direction: higher is better
inspect upstream source ->