The Leaderboard

The Leaderboard

Every edition, The AI Insider ranks the twenty models that actually matter — for real work, by real people — with reasons, not vibes.

Updated September 13, 2026 · Introductory Edition · 20 models ranked

---

1. 01

   Claude Fable 5.1 Max Effort

   Anthropic

   S NEW

   Top of the board.

   Top overall performance — advanced reasoning and the benchmark scores to match.

2. 02

   Claude Fable 5 Max Effort

   Anthropic

   S NEW

   The coder's pick.

   Exceptional real-world agentic coding and precision under load.

3. 03

   GPT-6 Astra Max Effort

   OpenAI

   S NEW

   Reasoning and maths, out front.

   Dominates general reasoning and mathematics.

4. 04

   Muse Spark 1.3 xHigh Effort

   Meta

   A NEW

   Throughput with a brain.

   High throughput paired with a strong intelligence index.

5. 05

   GPT-5.6 Sol Max Effort

   OpenAI

   A NEW

   Best in the browser and the terminal.

   Outstanding browsing and terminal-use execution.

6. 06

   GPT-5.5 Thinking xHigh Effort

   OpenAI

   A NEW

   Deep thinking, kept balanced.

   Balanced deep thinking and data analysis.

7. 07

   Claude 5 Opus Thinking Max Effort

   Anthropic

   A NEW

   Front-end's favourite.

   Exceptional front-end design and day-to-day coding utility.

8. 08

   Kimi K3

   Moonshot AI

   A NEW

   The open-weight standard-bearer.

   The leading open-weight alternative for complex reasoning.

9. 09

   Gemini 3.7 Flash High

   Google

   A NEW

   Fast, and cheap per token.

   High speed and efficient token performance.

10. 10

    Qwen 3.8 Max

    Alibaba

    B NEW

    Open weights that can do the maths.

    Strong open-weights reasoning and mathematics performance.

11. 11

    Grok 4.6

    xAI

    B NEW

    Long context, quick answers.

    High-context reasoning with fast output.

12. 12

    GPT-5.4 Thinking xHigh Effort

    OpenAI

    B NEW

    Reliable at maths and instructions.

    Strong maths and general instruction-following.

13. 13

    Muse Spark 1.2 xHigh Effort

    Meta

    B NEW

    Quick, and steady with tools.

    Fast processing speed with reliable multi-step tool use.

14. 14

    GPT-5.6 Terra Max Effort

    OpenAI

    B NEW

    Speed with analysis attached.

    High-speed generation with solid analytical skills.

15. 15

    DeepSeek V4 Pro 0813

    DeepSeek

    B NEW

    Frontier work at a discount.

    Cost-effective frontier performance.

16. 16

    Gemini 3.1 Pro Preview High

    Google

    C NEW

    The big-context veteran.

    Strong multi-modality and a large context window.

17. 17

    DeepSeek V4 Flash Vision Exp

    DeepSeek

    C NEW

    Screenshots to code, efficiently.

    Excellent vision-to-code efficiency.

18. 18

    Qwen 3.8 Flash Next

    Alibaba

    C NEW

    Built to run local, fast.

    Ultra-fast response times for local and open-weight scaling.

19. 19

    GLM-5.3

    Zhipu AI

    C NEW

    Made for agent workflows.

    High performance in agentic workflows.

20. 20

    Claude Sonnet 5

    Anthropic

    C NEW

    Smart, and sensibly priced.

    High GPQA reasoning scores paired with moderate task cost.

Scored on capability, reliability, price, and how each one behaves in daily use once the demo is over. Frontier-only: fine-tunes and wrappers don't count.