Inference
LiveBench

LiveBench Instruction Following

Mean score over paraphrasing, simplifying, summarising and story generation, each marked on whether the stated constraints were actually met.

ModelsModels with a result.55
Top resultBest result on this test.81.4%Gemini 3.8 Flash
Top-3 spreadPoints from first to third.3.4 pts
UpdatedDate the source changed the results.25 Jun 2026

Result and cost

020406080100$0.0003$0.001$0.003$0.01$0.03$0.1$0.3$1$3RESULTUS DOLLARS PER TASK · LOG SCALE

Results

55 results
#ModelResultThe score from the source.CostUS dollars to run one task.
01Gemini 3.8 Flashhigh effort81.4%$0.169
02Gemini 3.7 Flashhigh effort79.9%$0.097
03Muse Spark 1.3xhigh effort78.0%$0.133
04Qwen3.8-Flash-Next77.1%$0.018
05Claude Fable 5max effort75.8%$0.924
06Gemini 3.5 Flashhigh effort75.6%$0.133
07GPT-6 Astra Max75.6%$1.24
08Gemini 3.6 Flashhigh effort75.4%$0.115
09Grok 4.7xhigh effort75.3%$0.237
10Muse Spark 1.2xhigh effort74.3%$0.072
11Qwen3.8 Max74.1%$0.082
12Qwen3.7 Max74.0%$0.079
13Smaug Mini73.9%$0.014
14Claude Fable 5.1max effort73.0%$1.2
15Qwen3.8-27B72.7%$0.013
16Claude Opus 4.8max effort72.0%$0.471
17Grok 4.671.9%$0.044
18GPT-5.6 Sol Max71.8%$0.418
19Grok 4.571.5%$0.047
20Smaug Agentic71.0%$0.105
21DeepSeek V4 Flash Vision Exp71.0%$0.036
22GPT-5.5xhigh effort70.7%$0.227
23GPT-5.4xhigh effort70.2%$0.259
24Inklingxhigh effort70.1%$0.067
25DeepSeek V4.1 Flash Max70.0%$0.016
26Muse Spark 1.1xhigh effort69.6%$0.047
27GLM-5.369.3%$0.271
28Smaug Flash69.3%<$0.01
29Gemini 3.5 Flash-Litehigh effort67.2%$0.023
30GPT-5.4 nanoxhigh effort67.2%$0.045
31Claude Opus 4.7xhigh effort66.7%$0.281
32GPT-5.2-Codex66.4%$0.071
33Claude Opus 5.5max effort65.7%$0.937
34Grok Build 0.165.2%<$0.01
35DeepSeek V4 Pro 042365.0%$0.02
36GPT-5.6 Terra Max64.6%$0.332
37Kimi K2.6 Thinking64.4%$0.067
38DeepSeek V4 Flash 042364.3%<$0.01
39Claude Sonnet 5xhigh effort63.9%$0.259
40Claude Opus 5max effort63.8%$0.328
41Claude Opus 4.6 Thinking Autohigh effort63.3%$0.335
42Claude Sonnet 4.6 Thinking Automedium effort63.2%$0.248
43Grok 4.362.8%$0.027
44Claude Opus 4.5 Thinking 64Khigh effort62.5%$0.4
45GLM-5.262.3%$0.062
46GPT-5.2high effort61.8%$0.102
47Ox Alpha Max60.3%$0
48GPT-5.6 Luna Max60.1%$0.153
49GPT-5.4 minixhigh effort59.8%$0.172
50Union Alpha59.5%$0
51Qwen3.6 Plus58.3%$0.05
52MiniMax M357.5%$0.019
53Kimi K2.7 Code56.3%$0.047
54Qwen3.6-27B53.2%$0.046
55GLM-5.3-Flash52.8%$0.019