Published: June 14, 2026 | Last updated: June 14, 2026
TL;DR. Anthropic released Claude Fable 5 on June 9, 2026, calling it its most powerful public model ever. On the DataCore AI Leaderboard, which tracks 527 models from 58 providers in real time, Fable 5 does take the number 1 spot with a composite score of 63.9, ahead of GPT-5.5 (59.8) and Claude Opus 4.8 (59.7). But "best" depends on what you measure: GPT-5.2 leads on math (99), Mercury 2 is the fastest (947 tokens/sec), and 220 open-weight models are closing the gap. See the live board at airank.datacore.vn.
Did Claude Fable 5 actually take the number 1 spot?
Yes. As of June 14, 2026, Claude Fable 5 (Anthropic) sits at the top of the DataCore AI Leaderboard with a composite score of 63.9, built from an Intelligence Index of 64.9 and a coding score of 62. Every new flagship arrives with a claim to be "the best AI model." The difference is that you no longer have to take the launch blog post at face value. The board updates from live benchmark feeds roughly every 10 minutes, so the ranking you see is the ranking as it stands right now, not a screenshot from launch day.
| Rank | Model | Provider | Composite | Intelligence | Coding |
|---|---|---|---|---|---|
| 1 | Claude Fable 5 | Anthropic | 63.9 | 64.9 | 62 |
| 2 | GPT-5.5 (xhigh) | OpenAI | 59.8 | 60.2 | 59.1 |
| 3 | Claude Opus 4.8 | Anthropic | 59.7 | 61.4 | 56.7 |
| 4 | GPT-5.2 (xhigh) | OpenAI | 57.7 | 51.3 | 48.7 |
| 6 | Gemini 3.1 Pro Preview | 56.6 | 57.2 | 55.5 |
Takeaway: Fable 5 leads, but the gap to the next tier is small (about 4 points to GPT-5.5). This is a three-way frontier race between Anthropic, OpenAI, and Google, not a runaway.
What is Claude Fable 5, and why the hype?
Claude Fable 5 is Anthropic's first Mythos-class model made available to the general public, released on June 9, 2026 alongside the restricted Claude Mythos 5. Anthropic describes it as state-of-the-art across nearly all tested benchmarks, with standout results in software engineering, scientific research, vision, and autonomous task execution, scoring more than 10% higher than Claude Opus 4.8 on some evaluations (source: Anthropic, June 9, 2026). In high-risk domains such as cybersecurity, biology, and chemistry, the model deliberately blocks responses and falls back to Claude Opus 4.8.
On pricing: Fable 5 is free for Pro, Max, Team, and seat-based Enterprise plans through June 22, 2026. From June 23, 2026 it is priced at 10 USD per 1M input tokens and 50 USD per 1M output tokens (source: Anthropic, June 9, 2026). That free window is exactly why this week is the moment to compare it against the field.
"Best" depends on what you measure: speed, price, and the open-source surge
A single number hides the trade-offs that actually decide which model you should use. On the DataCore board right now:
- Best for math: GPT-5.2 (xhigh) at a math score of 99.
- Fastest output: Mercury 2 at roughly 947 tokens per second.
- Cheapest frontier model: Claude Opus 4.8 at around 10 USD per 1M output tokens.
- Longest time horizon: Claude Opus 4.6 at about 14 hours (METR's measure of how long a task a model can complete).
- Largest context window: Llama 4 Scout at 10M tokens.
The board also tracks 220 open-weight models out of 527 total. Chinese frontier models such as Qwen3.7-Max, DeepSeek V4, Kimi K2, and GLM cluster in the 50 to 57 Intelligence range, close enough to the Western frontier that, for high-volume workloads, cost becomes the deciding factor rather than raw capability. You can filter the board by open weights, price ceiling, minimum intelligence, and context size. Compare any two models side by side on the compare page.
How the DataCore AI Leaderboard works
The leaderboard aggregates four public sources, each measuring a different dimension: Artificial Analysis (Intelligence Index v4, coding, math, speed, price), METR (50% task-completion time horizon), Aider Polyglot (multi-language code-editing pass rate), and OpenRouter (model catalog, context window, modalities, live pricing). Each metric carries a colored dot showing its source. The full method is documented on the methodology page. No benchmark is invented, and nothing on the board is used for training.
Why this matters for Vietnam
For Vietnamese teams building with AI, the real question is rarely "which model is smartest in the abstract" but "which model gives me the capability I need at a cost I can defend." A live, source-attributed board lets a Vietnamese developer or business compare a frontier model against a far cheaper open-weight alternative before committing budget. That same discipline (structured, sourced, comparable data) is what DataCore brings to Vietnamese market data through its six structured data domains: Organization, Market, Economy, People, Location, and Media. Explore the data platform at datacore.vn.
See the live ranking of all 527 AI models.
airank.datacore.vn, intelligence, coding, math, speed and cost, updated in real time.
Frequently asked questions
What is the best AI model right now?
By the DataCore composite score, Claude Fable 5 (Anthropic) currently leads at 63.9, ahead of GPT-5.5 (59.8) and Claude Opus 4.8 (59.7), as of June 14, 2026. The ranking updates continuously at airank.datacore.vn.
Is Claude Fable 5 free?
It is free for Pro, Max, Team, and seat-based Enterprise plans through June 22, 2026. From June 23, 2026, it is priced at 10 USD per 1M input tokens and 50 USD per 1M output tokens (source: Anthropic).
Which AI model is the cheapest?
Among frontier models, Claude Opus 4.8 is one of the lowest cost at around 10 USD per 1M output tokens. The board tracks far cheaper open-weight models too.
What is the fastest LLM?
Mercury 2 reaches the highest output speed on the board at roughly 947 tokens per second, as measured by Artificial Analysis.
Where does the leaderboard data come from?
From Artificial Analysis, METR, OpenRouter, and Aider Polyglot. The data refreshes roughly every 10 minutes. Full detail is on the methodology page.
Để lại một bình luận
You must be logged in to post a comment.