Skip to main content
TokenCost logoTokenCost

The Best LLMs on SWE-bench Verified

Which model solves the most SWE-bench Verified issues?

Only 9 of the 152 buyable models publish a SWE-bench Verified number at all. The other 143 are named in full below, with whatever they publish instead.

Method: Published SWE-bench Verified pass@1, descending, exactly as reported. Every figure here comes from the provider's own announcement, because no independent lab publishes this suite across the catalogue. Pricing as of October 2026. 1 row below carries a score measured at a non-default reasoning tier and is marked as such; every affected model is named on the hub. Read the full method.

RankModelSWE-bench VerifiedTerminal-BenchQualityReported byBlended $/1M
#1Claude Opus 5Anthropic96.086.192.0Unstated / mixed$10.00
#2Claude Opus 4.8Anthropic88.684.687.0Unstated / mixed$10.00
#3Claude Sonnet 5Anthropic· non-default tier85.280.583.3Unstated / mixed$4.00
#4Nex-N2-ProNex AGI80.867.875.6Unstated / mixed$1.00
#5MiMo-V2-ProXiaomi78.0not published78.0Unstated / mixed$1.50
#6Mistral Medium 3.5Mistral77.650.666.8Unstated / mixed$3.00
#7Muse Glimmer 30BMeta76.051.766.3Vendor-reported$0.637
#8Nemotron 3 Ultra 550BNVIDIA71.9not published71.9Unstated / mixed$1.00
#9Qwen3 Coder NextAlibaba70.638.257.6Unstated / mixed$0.290

The 143 models that publish no SWE-bench Verified score

This list is short because the suite is rarely reported, not because the catalogue is small. Every buyable model without a published SWE-bench Verified number is named below, with whatever it does publish instead. An absent score is a fact about what a provider chose to release, and none of these are estimated to fill the table out.

OpenAI

32
  • GPT Image 2 — nothing published · $11.25/1M
  • GPT Image 2.5 — nothing published · $11.25/1M
  • GPT-4.1 — GPQA 66.6 · tau2 47.1 · $3.50/1M
  • GPT-4.1 Mini — Terminal-Bench 2.1 10.1 · GPQA 66.4 · tau2 52.9 · $0.700/1M
  • GPT-4.1 Nano — Terminal-Bench 2.1 3.7 · GPQA 51.2 · tau2 17.3 · $0.175/1M
  • GPT-4o — GPQA 54.3 · tau2 25.1 · $4.38/1M
  • GPT-4o Mini — Terminal-Bench 2.1 5.6 · GPQA 42.6 · $0.263/1M
  • GPT-5 — GPQA 84.2 · tau2 86.5 · $3.44/1M
  • GPT-5 Mini — GPQA 80.3 · tau2 71.1 · $0.688/1M
  • GPT-5 Nano — GPQA 67.0 · tau2 30.4 · $0.138/1M
  • GPT-5.1 — Terminal-Bench 2.1 52.4 · GPQA 87.3 · tau2 81.9 · $3.44/1M
  • GPT-5.2 — GPQA 86.4 · tau2 74.3 · $4.81/1M
  • GPT-5.3 Codex — GPQA 91.5 · tau2 86.0 · $4.81/1M
  • GPT-5.4 — GPQA 87.1 · tau2 74.6 · $5.63/1M
  • GPT-5.4 Mini — GPQA 82.3 · tau2 36.5 · $1.69/1M
  • GPT-5.4 Nano — GPQA 76.1 · tau2 52.6 · $0.463/1M
  • GPT-5.4 Pro — nothing published · $67.50/1M
  • GPT-5.5 — Terminal-Bench 2.1 80.5 · GPQA 92.6 · tau2 91.8 · $11.25/1M
  • GPT-5.5 Pro — nothing published · $67.50/1M
  • GPT-5.6 Cyber — nothing published · $28.13/1M
  • GPT-5.6 Luna — Terminal-Bench 2.1 53.2 · GPQA 85.9 · $0.450/1M
  • GPT-5.6 Sol — Terminal-Bench 2.1 86.1 · GPQA 92.6 · tau2 81.0 · $8.00/1M
  • GPT-5.6 Terra — Terminal-Bench 2.1 72.3 · GPQA 87.2 · tau2 72.8 · $4.50/1M
  • GPT-6 Luna — nothing published · $0.200/1M
  • GPT-6 Sol — nothing published · $4.00/1M
  • GPT-6.1 Sol — AA index 48.0 · $4.00/1M
  • o1 — GPQA 74.7 · tau2 62.6 · $26.25/1M
  • o1 Pro — AA index 18.9 · $262.50/1M
  • o3 — GPQA 82.7 · tau2 80.7 · $3.50/1M
  • o3 Mini — GPQA 74.8 · tau2 28.7 · $1.93/1M
  • o3-pro — GPQA 84.5 · $35.00/1M
  • o4 Mini — GPQA 78.4 · tau2 55.6 · $1.93/1M

Google

21

Alibaba

15

Anthropic

11

Zhipu

9
  • GLM-4.7 — Terminal-Bench 2.1 45.3 · GPQA 85.9 · tau2 95.9 · $1.00/1M
  • GLM-4.7-FlashX — GPQA 58.1 · tau2 98.8 · $0.152/1M
  • GLM-5 — GPQA 82.0 · tau2 98.2 · $1.55/1M
  • GLM-5 Turbo — GPQA 84.7 · tau2 98.5 · $1.90/1M
  • GLM-5.1 — Terminal-Bench 2.1 61.8 · GPQA 86.8 · tau2 97.7 · $2.15/1M
  • GLM-5.2 — Terminal-Bench 2.1 77.9 · GPQA 89.5 · tau2 99.1 · $2.15/1M
  • GLM-5.3 — nothing published · $2.15/1M
  • GLM-5.3 Prime — nothing published · $4.30/1M
  • GLM-5.3-Flash — nothing published · $0.237/1M

Meta

8

Mistral

7

xAI

6
  • Grok 4.20 — GPQA 91.1 · tau2 93.0 · $1.56/1M
  • Grok 4.3 — GPQA 89.0 · tau2 91.2 · $1.56/1M
  • Grok 4.5 — Terminal-Bench 2.1 81.6 · GPQA 93.1 · $3.00/1M
  • Grok 4.6 — nothing published · $3.00/1M
  • Grok 4.7 — nothing published · $3.00/1M
  • Grok Build 0.1 — nothing published · $1.25/1M

MiniMax

4

Moonshot

4
  • Kimi K2.5 — Terminal-Bench 2.1 45.7 · GPQA 87.9 · tau2 95.9 · $1.20/1M
  • Kimi K2.6 — Terminal-Bench 2.1 65.9 · GPQA 91.1 · tau2 95.9 · $1.71/1M
  • Kimi K2.7 Code — Terminal-Bench 2.1 67.4 · GPQA 89.6 · tau2 90.1 · $1.71/1M
  • Kimi K3 — Terminal-Bench 2.1 85.0 · GPQA 93.5 · $6.00/1M

Xiaomi

3

Ant Group

2

ByteDance

2

Cohere

2
  • Command A — Terminal-Bench 2.1 22.8 · GPQA 76.1 · tau2 80.7 · $4.38/1M
  • Command A+ — Terminal-Bench 2.1 22.8 · GPQA 76.1 · tau2 80.7 · $4.38/1M

DeepSeek

2

Inception Labs

2

Kwaipilot

2

Sakana AI

2

Tencent

2
  • Hy3 — nothing published · $0.231/1M
  • Hy4 preview — nothing published · $1.25/1M

Amazon

1

Meituan

1
  • LongCat-2.0 — Terminal-Bench 2.1 50.2 · GPQA 78.0 · $1.30/1M

Microsoft

1

NVIDIA

1

Perplexity

1

Poolside

1

StepFun

1
  • Step 3.7 Flash — Terminal-Bench 2.1 39.3 · GPQA 80.9 · tau2 98.5 · $0.438/1M

Frequently asked

Nearby questions, answered elsewhere

Terminal-Bench 2.1
The suite most of the models below publish instead.
Both suites, weighted
The composite, which does not penalise a model for publishing only one of the two.

All rankings

The hub and the methodThe Best LLMs by Benchmark ScoreThe Best Value LLMs per DollarThe Best LLMs by Context WindowThe Cheapest LLMs With a 1M-Token Context WindowThe Best LLMs on Terminal-Bench 2.1The Best LLMs on Terminal-Bench HardThe Best LLMs on GPQA DiamondThe Best LLMs for Tool Use on tau2-bench