# Best AI Models in Kilo

> Live rankings of AI coding models from real Kilo Code usage across 5M+ developers. HTML: https://kilo.ai/leaderboard

Generated: 2026-09-17

## Methodology

Models earn their rank from the developers using them, not from a spec sheet. Rankings use real token usage across 500+ hosted models and update about every 5 minutes.

## Featured coding models

1. GPT-5.6 Sol — 76.2% KiloBench completion — A flagship model for long-horizon agents and complex problem solving
   Model: https://kilo.ai/models/openai-gpt-5-6-sol
   Read more: https://blog.kilo.ai/p/the-new-gpt-56-models-will-blow-your

2. Claude Opus 4.8 — 74.3 Coding Index — Built for complex planning, orchestration, and long-running professional tasks
   Model: https://kilo.ai/models/anthropic-claude-opus-4-8
   Read more: https://blog.kilo.ai/p/we-audited-the-same-codebase-with

3. Grok 4.6 — 73.0% KiloBench completion — Frontier coding and STEM performance tuned for long-running agents
   Model: https://kilo.ai/models/x-ai-grok-4-6
   Read more: https://x.ai/news/grok-4-6

4. Step 3.7 Flash — 256k context window — A fast, open-weight model designed for efficient agentic workflows
   Read more: https://blog.kilo.ai/p/new-models-from-stepfun-and-minimax

## KiloBench (Terminal Bench 2.0)

Cost vs performance across the best coding models. Source: https://www.tbench.ai/

| Rank | Model | Completion | Cost per attempt | URL |
| --- | --- | --- | --- | --- |
| 1 | GPT-6 Astra ($$$$) | 79.3% | $107.29 | https://kilo.ai/models/openai-gpt-6-astra |
| 2 | GPT-5.6 Sol | 76.2% | $87.41 | https://kilo.ai/models/openai-gpt-5-6-sol |
| 3 | Claude Fable 5.1 ($$$$) | 76.2% | $91.41 | https://kilo.ai/models/anthropic-claude-fable-5-1 |
| 4 | DeepSeek V4.1 Flash | 75.3% | $2.58 | https://kilo.ai/models/deepseek-deepseek-v4-1-flash |
| 5 | Gemini 3.8 Flash (50% off) | 75.3% | $112.27 | https://kilo.ai/models/google-gemini-3-8-flash |
| 6 | GPT-5.5 | 74.2% | $72.63 | https://kilo.ai/models/openai-gpt-5-5 |
| 7 | Grok 4.6 | 73.0% | $33.83 | https://kilo.ai/models/x-ai-grok-4-6 |
| 8 | Kimi K3 | 72.8% | $48.38 | https://kilo.ai/models/moonshotai-kimi-k3 |
| 9 | Claude Opus 5 | 71.5% | $113.54 | https://kilo.ai/models/anthropic-claude-opus-5 |
| 10 | Claude Fable 5 ($$$$) | 71.0% | $87.52 | https://kilo.ai/models/anthropic-claude-fable-5 |

## Top models by mode

### Code

| Rank | Model | Share | URL |
| --- | --- | --- | --- |
| 1 | step-3.7-flash | 22.6% |  |
| 2 | Poolside: Laguna S 2.1 (free) | 19.2% | https://kilo.ai/models/poolside-laguna-s-2-1-free |
| 3 | NVIDIA: Nemotron 3 Ultra | 13.0% | https://kilo.ai/models/nvidia-nemotron-3-ultra-550b-a55b |
| 4 | Dots Studio: Dots3-Note Preview (free) | 8.5% | https://kilo.ai/models/dots-studio-dots-3-note-preview-free |
| 5 | inclusionAI: Ling 3.0 Flash VL (free) | 7.7% | https://kilo.ai/models/inclusionai-ling-3-0-flash-vl-free |
| 6 | Nex AGI: Nex-N2.5-Pro (free) | 5.9% | https://kilo.ai/models/nex-agi-nex-n2-5-pro-free |
| 7 | DeepSeek: DeepSeek V4.1 Flash | 4.9% | https://kilo.ai/models/deepseek-deepseek-v4-1-flash |
| 8 | Z.ai: GLM 5.3 Flash | 1.7% | https://kilo.ai/models/z-ai-glm-5-3-flash |
| 9 | MiniMax: MiniMax M3 | 1.4% | https://kilo.ai/models/minimax-minimax-m3 |
| 10 | NVIDIA: Nemotron 3 Super (free) | 1.2% | https://kilo.ai/models/nvidia-nemotron-3-super-120b-a12b-free |

### Plan

| Rank | Model | Share | URL |
| --- | --- | --- | --- |
| 1 | Poolside: Laguna S 2.1 (free) | 16.5% | https://kilo.ai/models/poolside-laguna-s-2-1-free |
| 2 | step-3.7-flash | 12.9% |  |
| 3 | NVIDIA: Nemotron 3 Ultra | 11.2% | https://kilo.ai/models/nvidia-nemotron-3-ultra-550b-a55b |
| 4 | DeepSeek: DeepSeek V4.1 Flash | 9.2% | https://kilo.ai/models/deepseek-deepseek-v4-1-flash |
| 5 | Dots Studio: Dots3-Note Preview (free) | 7.7% | https://kilo.ai/models/dots-studio-dots-3-note-preview-free |
| 6 | Nex AGI: Nex-N2.5-Pro (free) | 7.5% | https://kilo.ai/models/nex-agi-nex-n2-5-pro-free |
| 7 | MiniMax: MiniMax M3 | 5.7% | https://kilo.ai/models/minimax-minimax-m3 |
| 8 | inclusionAI: Ling 3.0 Flash VL (free) | 5.4% | https://kilo.ai/models/inclusionai-ling-3-0-flash-vl-free |
| 9 | Z.ai: GLM 5.3 Flash | 2.1% | https://kilo.ai/models/z-ai-glm-5-3-flash |
| 10 | MoonshotAI: Kimi K3 | 1.7% | https://kilo.ai/models/moonshotai-kimi-k3 |

### Ask

| Rank | Model | Share | URL |
| --- | --- | --- | --- |
| 1 | Poolside: Laguna S 2.1 (free) | 17.3% | https://kilo.ai/models/poolside-laguna-s-2-1-free |
| 2 | step-3.7-flash | 17.3% |  |
| 3 | NVIDIA: Nemotron 3 Ultra | 11.3% | https://kilo.ai/models/nvidia-nemotron-3-ultra-550b-a55b |
| 4 | Dots Studio: Dots3-Note Preview (free) | 9.7% | https://kilo.ai/models/dots-studio-dots-3-note-preview-free |
| 5 | Nex AGI: Nex-N2.5-Pro (free) | 6.8% | https://kilo.ai/models/nex-agi-nex-n2-5-pro-free |
| 6 | inclusionAI: Ling 3.0 Flash VL (free) | 6.7% | https://kilo.ai/models/inclusionai-ling-3-0-flash-vl-free |
| 7 | DeepSeek: DeepSeek V4.1 Flash | 2.2% | https://kilo.ai/models/deepseek-deepseek-v4-1-flash |
| 8 | MoonshotAI: Kimi K3 | 1.9% | https://kilo.ai/models/moonshotai-kimi-k3 |
| 9 | NVIDIA: Nemotron 3 Super (free) | 1.8% | https://kilo.ai/models/nvidia-nemotron-3-super-120b-a12b-free |
| 10 | Z.ai: GLM 5.3 Flash | 1.8% | https://kilo.ai/models/z-ai-glm-5-3-flash |

### Debug

| Rank | Model | Share | URL |
| --- | --- | --- | --- |
| 1 | step-3.7-flash | 29.0% |  |
| 2 | Poolside: Laguna S 2.1 (free) | 18.6% | https://kilo.ai/models/poolside-laguna-s-2-1-free |
| 3 | NVIDIA: Nemotron 3 Ultra | 13.3% | https://kilo.ai/models/nvidia-nemotron-3-ultra-550b-a55b |
| 4 | Dots Studio: Dots3-Note Preview (free) | 7.3% | https://kilo.ai/models/dots-studio-dots-3-note-preview-free |
| 5 | Nex AGI: Nex-N2.5-Pro (free) | 5.9% | https://kilo.ai/models/nex-agi-nex-n2-5-pro-free |
| 6 | inclusionAI: Ling 3.0 Flash VL (free) | 5.0% | https://kilo.ai/models/inclusionai-ling-3-0-flash-vl-free |
| 7 | Z.ai: GLM 5.3 Flash | 3.0% | https://kilo.ai/models/z-ai-glm-5-3-flash |
| 8 | Anthropic: Claude Opus 5 | 2.3% | https://kilo.ai/models/anthropic-claude-opus-5 |
| 9 | DeepSeek: DeepSeek V4.1 Flash | 1.9% | https://kilo.ai/models/deepseek-deepseek-v4-1-flash |
| 10 | free | 1.8% |  |

### Review

| Rank | Model | Share | URL |
| --- | --- | --- | --- |
| 1 | DeepSeek: DeepSeek V4.1 Flash | 49.1% | https://kilo.ai/models/deepseek-deepseek-v4-1-flash |
| 2 | step-3.7-flash | 9.8% |  |
| 3 | MoonshotAI: Kimi K3 | 5.2% | https://kilo.ai/models/moonshotai-kimi-k3 |
| 4 | Nex AGI: Nex-N2.5-Pro (free) | 4.9% | https://kilo.ai/models/nex-agi-nex-n2-5-pro-free |
| 5 | Poolside: Laguna S 2.1 (free) | 4.4% | https://kilo.ai/models/poolside-laguna-s-2-1-free |
| 6 | NVIDIA: Nemotron 3 Ultra | 3.5% | https://kilo.ai/models/nvidia-nemotron-3-ultra-550b-a55b |
| 7 | Anthropic: Claude Sonnet 5 | 3.3% | https://kilo.ai/models/anthropic-claude-sonnet-5 |
| 8 | Dots Studio: Dots3-Note Preview (free) | 3.3% | https://kilo.ai/models/dots-studio-dots-3-note-preview-free |
| 9 | gpt-5.6-luna-pro | 2.7% |  |
| 10 | OpenAI: GPT-6 Astra ($$$$) | 1.6% | https://kilo.ai/models/openai-gpt-6-astra |

### kiloclaw

| Rank | Model | Share | URL |
| --- | --- | --- | --- |
| 1 | deepseek-v4-flash-0731 | 16.1% |  |
| 2 | MiniMax: MiniMax M3 | 14.8% | https://kilo.ai/models/minimax-minimax-m3 |
| 3 | inclusionAI: Ling 3.0 Flash VL (free) | 13.4% | https://kilo.ai/models/inclusionai-ling-3-0-flash-vl-free |
| 4 | Z.ai: GLM 5.3 Flash | 9.6% | https://kilo.ai/models/z-ai-glm-5-3-flash |
| 5 | Dots Studio: Dots3-Note Preview (free) | 7.2% | https://kilo.ai/models/dots-studio-dots-3-note-preview-free |
| 6 | NVIDIA: Nemotron 3 Ultra | 5.9% | https://kilo.ai/models/nvidia-nemotron-3-ultra-550b-a55b |
| 7 | Anthropic: Claude Opus 5 | 5.6% | https://kilo.ai/models/anthropic-claude-opus-5 |
| 8 | Nex AGI: Nex-N2.5-Pro (free) | 5.6% | https://kilo.ai/models/nex-agi-nex-n2-5-pro-free |
| 9 | Poolside: Laguna S 2.1 (free) | 2.7% | https://kilo.ai/models/poolside-laguna-s-2-1-free |
| 10 | DeepSeek: DeepSeek V4 Flash 0423 | 2.5% | https://kilo.ai/models/deepseek-deepseek-v4-flash |

## Top models today

Most-used models across Kilo Code in the last 24 hours.

| Rank | Model | Tokens (24h) | URL |
| --- | --- | --- | --- |
| 1 | step-3.7-flash | 119.0B |  |
| 2 | Poolside: Laguna S 2.1 (free) | 94.9B | https://kilo.ai/models/poolside-laguna-s-2-1-free |
| 3 | Dots Studio: Dots3-Note Preview (free) | 75.7B | https://kilo.ai/models/dots-studio-dots-3-note-preview-free |
| 4 | inclusionAI: Ling 3.0 Flash VL (free) | 64.2B | https://kilo.ai/models/inclusionai-ling-3-0-flash-vl-free |
| 5 | NVIDIA: Nemotron 3 Ultra | 64.0B | https://kilo.ai/models/nvidia-nemotron-3-ultra-550b-a55b |
| 6 | DeepSeek: DeepSeek V4.1 Flash | 51.9B | https://kilo.ai/models/deepseek-deepseek-v4-1-flash |
| 7 | Nex AGI: Nex-N2.5-Pro (free) | 50.8B | https://kilo.ai/models/nex-agi-nex-n2-5-pro-free |
| 8 | free | 13.5B |  |
| 9 | ling-3.0-flash-fin | 13.0B |  |
| 10 | Z.ai: GLM 5.3 Flash | 12.9B | https://kilo.ai/models/z-ai-glm-5-3-flash |

## How to read the leaderboard

1. Start with usage — The rankings reflect real token usage by Kilo Code developers, not synthetic benchmarks.
2. Filter by mode — Use the Top Models by Mode section to see which models lead in Code, Plan, Debug, Ask, and Orchestrator.
3. Open the model page — Each model links to a dedicated page with benchmark scores, pricing, context length, and speed data.
4. Switch in Kilo Code — All 500+ models are available in Kilo Code. Switch from the model selector at any time.

## FAQ

### What is the Kilo Code AI Model Leaderboard?

The Kilo Code leaderboard shows live rankings of AI coding models based on real token usage by 5M+ developers. Rankings reflect genuine developer preference and update every 5 minutes.

### How are AI coding models ranked?

Models are ranked by total token usage from Kilo Code developers. The ranking reflects real-world developer preference and can be filtered by modes such as Code, Plan, Debug, Ask, and Orchestrator.

### Which AI model is best for coding?

The current top-ranked model is step-3.7-flash, based on real developer usage. The best AI model for your workflow can vary by coding, planning, debugging, and agent tasks, so compare the live rankings by mode before choosing.

### How often does the leaderboard update?

The leaderboard updates every 5 minutes with fresh usage data from real Kilo Code developers.

## Related

- Compare models: https://kilo.ai/models/compare
- Open source models: https://kilo.ai/open-source-models
- Browse by lab: https://kilo.ai/models/by
- KiloBench: https://kilo.ai/kilobench
- Inference: https://kilo.ai/leaderboard/inference
- Provider race: https://kilo.ai/leaderboard/race
- PinchBench: https://pinchbench.com
- HTML leaderboard: https://kilo.ai/leaderboard
- This file: https://kilo.ai/leaderboard/llm.txt