Best AI Models for Coding

Coding and agentic models ordered by their published coding pillar, combining the benchmark variants available in this snapshot.

updated Oct 9, 2026 · 15 models ranked

How this list is ranked: Ranked by the coding pillar. Its benchmark definitions and within-pillar weights come from the current snapshot below; model pages show the harness, variant and source for each result.

Current inputs: aider polyglot (weight 1); livebench coding code completion 2026_06_25 (weight 1); livebench coding code generation 2026_06_25 (weight 1); livebench coding javascript 2026_06_25 (weight 1); livebench coding python 2026_06_25 (weight 1); livebench coding typescript 2026_06_25 (weight 1); swe bench pro (weight 1); swe bench pro public (weight 1); swe bench pro qwen refined and corrected task set (weight 1); swe bench pro system card 2026 06 (weight 1); swe bench verified (weight 1); terminal bench (weight 1); terminal bench v0 1 (weight 1); terminal bench v2 0 (weight 1); terminal bench v2 1 (weight 1); terminal bench v3 0 (weight 1); terminal bench v4 0 (weight 1). Pillar scores can have different evidence coverage; review the individual results.

#	Model	Coding pillar	SI Score	Confidence	Price in	Context
1	Fugu Ultra sakana provisional	77.9	57.9	100% confidence 100 percent, Full	—	1M
2	Qwen3.6 Max Preview Alibaba / Qwen provisional	76.7	66.4	64% confidence 64 percent, Medium	$1.30	262K
3	Qwen3.5 397B-A17B Alibaba / Qwen provisional	76.4	64.0	69% confidence 69 percent, Medium	$0.17	262K
4	MiniMax-M2.5 minimax provisional	75.8	57.8	100% confidence 100 percent, Full	$0.30	205K
5	GLM-5.1 Z.ai provisional	74.2	63.3	64% confidence 64 percent, Medium	$1.40	200K
6	Claude Opus 4.1 Anthropic provisional	73.3	51.5	80% confidence 80 percent, High	—	200K
7	DeepSeek V4.1 Flash DeepSeek	73.3	66.2	69% confidence 69 percent, Medium	$0.15	1M
8	Claude Opus 5.5 Anthropic	72.5	78.4	100% confidence 100 percent, Full	$4.00	1M
9	GLM-5 Z.ai provisional	72.4	59.4	90% confidence 90 percent, High	$1.00	205K
10	Kimi K2.5 Moonshot AI provisional	72.4	59.0	100% confidence 100 percent, Full	$0.60	262K
11	Claude Sonnet 4.5 Anthropic provisional	71.3	53.9	80% confidence 80 percent, High	$3.00	200K
12	Kimi K3 Moonshot AI provisional	71.1	71.0	100% confidence 100 percent, Full	$3.00	1M
13	Claude Opus 4 Anthropic provisional	70.6	55.5	100% confidence 100 percent, Full	$15.00	200K
14	Claude Fable 5 Anthropic provisional	70.2	76.8	100% confidence 100 percent, Full	$10.00	1M
15	DeepSeek V3.2 DeepSeek provisional	70.0	53.3	56% confidence 56 percent, Medium	$0.28	128K

Task pages rank on one transparent metric, not the blended SI Score — see methodology for how pillars and confidence are computed. Missing values mean the source has not reported them.