![[AI Computing Power Rankings] How does the world champion from six months ago rank today? Claude Opus 5.5 remains at the top.](https://thebaseprinciple.com/uploads/cover_153bf9c8b1.jpg)
This week’s two champions: Intelligence ranking Claude Opus 5.5 (Max, Default Fallback) (57.6 points), Coding ranking Claude Fable 5.1 (Max, Default Fallback) (81.6 points). See the table for the rise and fall indicators.
With a perfect score of 100, don’t be fooled by the numbers: the test includes benchmarks like Humanity’s Last Exam, which are extremely difficult. A score of 57.6 is already the highest in human history.
Top Labs: Current Status + Two Metrics
3/6 months ago is the baseline; this week +1 month ago is the current status, indicating who is stagnant and who has made a leap.
Intelligence
| Lab | Current Champion | 1 Month Ago | 3 Months Ago | 6 Months Ago |
|---|---|---|---|---|
| Overall Champion | None | Claude Opus 5.5, 57.6 | Claude Opus 5, 49.7 | GPT-5.5, 38.4 |
| Anthropic | None | Claude Opus 5.5, 57.6 | Claude Opus 5, 49.7 | Claude Opus 4.7, 30.9 |
| OpenAI | None | GPT-6 Astra, 52.7 | GPT-5.5, 38.4 | Same as left |
| None | Gemini 3.5 Flash, 32.6 | Same as left | Gemini 2.5 Pro, 16.1 | |
| xAI | None | Grok 4.7, 46.3 | Grok Build 0.1 0616, 27.2 | Grok 4.20 0309, 25.2 |
| Qwen | None | Qwen 3.5 27B, 22.9 | Same as left | Same as left |
| 智譜 | None | GLM-5.3-Flash, 41.8 | GLM-5, 27.9 | Same as left |
Coding
| Lab | Current Champion | 1 Month Ago | 3 Months Ago | 6 Months Ago |
|---|---|---|---|---|
| Overall Champion | None | Claude Fable 5.1, 81.6 | Claude Opus 5, 77.0 | GPT-5.5, 74.9 |
| Anthropic | None | Claude Fable 5.1, 81.6 | Claude Opus 5, 77.0 | Claude 3.7 Sonnet, 36.4 |
| OpenAI | None | GPT-6 Astra, 76.9 | GPT-5.5, 74.9 | Same as left |
| None | Gemini 3.5 Flash, 70.1 | Same as left | Gemini 2.5 Pro, 33.3 | |
| xAI | None | Grok 4.6, 74.4 | Grok Build 0.1 0616, 51.5 | None |
| Qwen | None | Qwen 3.5 397B A17B, 48.2 | Same as left | Same as left |
| 智譜 | None | GLM-5.3-Flash, 71.5 | None | Same as left |
Not surpassing the champion from six months ago means falling behind for half a year; only surpassing the champion from three months ago means catching up with the times.
Crossing the Line: Which Models Have Caught Up? Which Are Free to Use?
★ = Surpassed the 3-month strong line; ✅ = This version is free; ◐ = Free version of the same series; — = None. Free versions are mostly time-limited, based on the date of publication.
Intelligence — Surpassed 6-Month Line (38.4 points) 21 Models
| Model | Score | Strong Line | OpenRouter Free | Opencode Free |
|---|---|---|---|---|
| Claude Opus 5.5 (Max, Default Fallback) (—) | 57.6 | ★ | — | — |
| Claude Fable 5.1 (Max, Default Fallback) (—) | 53.4 | ★ | — | — |
| GPT-6 Astra (Max) (—) | 52.7 | ★ | — | — |
| GPT-6 Astra (Xhigh) (—) | 52.4 | ★ | — | — |
| GPT-6.1 Sol (Xhigh) (—) | 51.0 | ★ | — | — |
| GPT-6.1 Sol (High) (—) | 50.2 | ★ | — | — |
| Claude Opus 5 (Xhigh) (—) | 49.7 | Strong line itself | — | — |
| Claude Fable 5.1 (Medium, Default Fallback) (—) | 48.9 | — | — | |
| Claude Fable 5.1 (Low, Default Fallback) (—) | 46.8 | — | — | |
| Claude Sonnet 5.5 (High, Default Fallback) (—) | 46.8 | — | — | |
| MiMo-V2.6-Pro | 46.3 | ◐ (same series) | ◐ (same series) | |
| Grok 4.7 (High) | 46.3 | — | — | |
| Step 5 Preview | 43.7 | — | — | |
| Grok 4.6 (Medium) (—) | 42.8 | — | — | |
| Grok 4.7 (Low) | 42.2 | — | — | |
| GLM-5.3-Flash (—) | 41.8 | ◐ (same series) | ◐ (same series) | |
| Claude Haiku 5.5 (Xhigh) | 41.2 | — | — (originally US$0.1 extremely cheap) | |
| Claude Sonnet 5.5 (Medium, Default Fallback) | 40.8 | — | — | |
| Muse Spark 1.2 (Xhigh) (—) | 39.6 | — | — | |
| Claude Opus 5 (Low) | 39.4 | — | — | |
| Muse Spark Open (Xhigh) | 38.9 | — | — |
Coding — Surpassed 6-Month Line (74.9 points) 6 Models
| Model | Score | Strong Line | OpenRouter Free | Opencode Free |
|---|---|---|---|---|
| Claude Fable 5.1 (Max, Default Fallback) (—) | 81.6 | ★ | — | — |
| Claude Fable 5.1 (Medium, Default Fallback) (—) | 77.1 | ★ | — | — |
| Claude Opus 5 (Xhigh) (—) | 77.0 | Strong line itself | — | — |
| GPT-6 Astra (Max) (—) | 76.9 | — | — | |
| GPT-6 Astra (Xhigh) (—) | 75.9 | — | — | |
| Claude Fable 5.1 (Low, Default Fallback) (—) | 75.2 | — | — |
Cost-Effectiveness: 6-Month Version vs 3-Month Version
6-month versions qualify if they exceed the sufficient line and surpass the strong line by 1.5 times; 3-month versions qualify if they exceed the strong line, with no additional multiplier for the same level.
6-Month Version
| Intelligence (Model/Score/Input Price/Value per Dollar) | Coding (Model/Score/Input Price/Value per Dollar) | |
|---|---|---|
| 1 | Claude Haiku 5.5 / 41.2 / US$0.1 / 412.0 | Claude Opus 5 / 77 / US$5 / 15.4 |
| 2 | GLM-5.3-Flash / 41.8 / US$0.15 / 278.7 | Claude Fable 5.1 / 81.6 / US$10 / 12.2 |
| 3 | MiMo-V2.6-Pro / 46.3 / US$0.435 / 106.4 | Claude Fable 5.1 / 77.1 / US$10 / 11.6 |
3-Month Version
| Intelligence (Model/Score/Input Price/Value per Dollar) | Coding (Model/Score/Input Price/Value per Dollar) | |
|---|---|---|
| 1 | GPT-6.1 Sol / 51 / US$2 / 25.5 | Claude Fable 5.1 / 81.6 / US$10 / 8.2 |
| 2 | GPT-6.1 Sol / 50.2 / US$2 / 25.1 | Claude Fable 5.1 / 77.1 / US$10 / 7.7 |
| 3 | Claude Opus 5.5 / 57.6 / US$4 / 14.4 | Missing |
Best Free Options + Best Value for Money
Free - 6-Month Version (All Exceed Sufficient Line)
Missing.
Free - 3-Month Version: Missing. None exceed the strong line, free options cannot keep up with the cutting edge, but they surpass the world champion from six months ago, sufficient for daily use.
Best Value for Money
| 6-Month Version Answer | 3-Month Version Answer | |
|---|---|---|
| Intelligence | Claude Haiku 5.5 (Xhigh) (41.2 points / US$0.1 / 412.0) | GPT-6.1 Sol (Xhigh) (51 points / US$2 / 25.5) |
| Coding | Claude Opus 5 (Xhigh) (77 points / US$5 / 15.4) | Claude Fable 5.1 (Max, Default Fallback) (81.6 points / US$10 / 8.2) |
Computing Power Timeline: What Chips Are Used to Train Current Models?
| Chip/Architecture | Announcement | Shipping Status |
|---|---|---|
| B200 (Blackwell) | March 2024 GTC | Mature for large-scale commercial use, major cloud providers |
| B300 (Blackwell Ultra) | End of 2024 / Early 2025 for specifications | Shipping ramping up, DGX B300 available for purchase, hourly rental in the cloud |
| Vera Rubin (R100 + HBM4) | 2025/2026 GTC finalized | Production starts mid-2026, partners begin deliveries in the second half of the year, not widely commercialized |
Pattern: From the start of shipping, it takes 9–12 months before the next wave of AI computing power upgrades peaks. B300 will only be widely deployed this year, with the next generation of models trained using it expected to be seen by the end of the year; products trained using Rubin will not be seen until 2027.
Other Updates
・Most Expensive: The most expensive input on the entire list, both costly and weak, is not recommended for use; current flagship price is US$10.
・Specialization King: Intelligence has never surpassed the line but is world-class in Coding, specializing in writing code.
Data Source: Artificial Analysis Intelligence Index and Coding Agent Index, publicly available free data subset; free model list can be found at OpenRouter and Opencode, limited-time free access may end at any time, all based on the date of publication.