The Inference Report

May 9, 2026

The Value Frontier

Price vs Intelligence — lower price, higher score is better.

Fast and Smart

Speed vs Intelligence — higher on both axes is better.

Efficiency

Speed vs Price — faster and cheaper is better.

SWE-rebench

Daily rankings from SWE-rebench, a benchmark designed to fairly compare LLM capabilities on real-world software engineering tasks. Unlike other evaluations, it uses a standardized scaffolding for all models, continuously updates its dataset to prevent contamination, and runs each model five times to account for stochastic variance.

#	Model	Score
1	Claude Opus 4.6	65.3%
2	gpt-5.2-2025-12-11-medium	64.4%
3	GLM-5	62.8%
4	Junie	62.8%
5	gpt-5.4-2026-03-05-medium	62.8%
6	GLM-5.1	62.7%
7	Gemini 3.1 Pro Preview	62.3%
8	DeepSeek-V3.2	60.9%
9	Claude Sonnet 4.6	60.7%
10	Claude Sonnet 4.5	60.0%

Intelligence Index — Leaderboard

Artificial Analysis composite index across coding, math, and reasoning benchmarks.

#	Model	Score	tok/s	$/1M
1	GPT-5.5	60.2	70	$11.25
2	Claude Opus 4.7	57.3	55	$10.94
3	Gemini 3.1 Pro Preview	57.2	120	$4.50
4	GPT-5.4	56.8	76	$5.63
5	Kimi K2.6	53.9	39	$1.71
6	MiMo-V2.5-Pro	53.8	61	$1.50
7	GPT-5.3 Codex	53.6	77	$4.81
8	Grok 4.3	53.2	82	$1.56
9	Claude Opus 4.6	53	50	$10.94
10	Muse Spark	52.1	0	$0.00

Speed — Leaderboard

Output tokens per second — higher is faster. Minimum intelligence score of 40.

#	Model	tok/s
1	Gemini 3 Flash Preview	198
2	Qwen3.6 35B A3B	189
3	GPT-5.4 mini	174
4	GPT-5.1 Codex	170
5	GPT-5 Codex	168
6	Qwen3.5 122B A10B	161
7	GPT-5.4 nano	154
8	MiMo-V2-Flash	139
9	GPT-5.1	134
10	Gemini 3.1 Pro Preview	120

Price — Leaderboard

Blended cost per 1M tokens (3:1 input/output) — lower is cheaper. Minimum intelligence score of 40.

#	Model	$/1M
1	MiMo-V2-Flash	$0.15
2	DeepSeek V4 Flash	$0.175
3	DeepSeek V3.2	$0.337
4	GPT-5.4 nano	$0.463
5	MiniMax-M2.7	$0.525
6	KAT Coder Pro V2	$0.525
7	MiniMax-M2.5	$0.525
8	Qwen3.6 35B A3B	$0.557
9	GPT-5 mini	$0.688
10	MiMo-V2.5	$0.72