AI ranking
Cost / performance
The ratio between overall performance and the public monthly subscription (Claude.ai Pro, ChatGPT Plus, Gemini Advanced…). API prices and per-token costs are not taken into account.
Visual top 10
- 01 Deepseek chat v3 (prev) 99.99
- 02 Grok 3 mini beta (high) 99.78
- 03 Grok 3 mini beta (low) 99.75
- 04 Gemini 2.5 Flash Preview 05 20 (no think) 99.56
- 05 Deepseek V3.2 Exp (reasoner) 99.47
- 06 Gemini 2.5 Flash Preview 04 17 (default) 99.18
- 07 Codestral 25.01 99.11
- 08 Claude opus 5.5 (high) 98.58
- 09 GPT 6 sol (daybreak blue, max) 98.50
- 10 Claude Fable 5 1 Max 97.52
Compare models. Tick up to 3 models in the table below, then run the comparison: who wins on which use, by how much, and how solid the measurement is.
| # | Model | Vendor | Score | Monthly subscription | Assessment | |
|---|---|---|---|---|---|---|
| 01 | DeepSeek | 99.99 / 100 | Free | Partial | ||
| 02 | xAI | 99.78 / 100 | Free | Partial | ||
| 03 | xAI | 99.75 / 100 | Free | Partial | ||
| 04 | 99.56 / 100 | €22/month | Partial | |||
| 05 | DeepSeek | 99.47 / 100 | Free | Partial | ||
| 06 | 99.18 / 100 | €22/month | Partial | |||
| 07 | Mistral | 99.11 / 100 | €15/month | Partial | ||
| 08 | Anthropic | 98.58 / 100 | Not available | Partial | ||
| 09 | OpenAI | 98.50 / 100 | €23/month | Partial | ||
| 10 | Anthropic | 97.52 / 100 | €23/month | Partial | ||
| 11 | Anthropic | 97.51 / 100 | €23/month | Partial | ||
| 12 | Anthropic | 97.50 / 100 | Not available | Partial | ||
| 13 | Anthropic | 97.16 / 100 | Not available | Partial | ||
| 14 | OpenAI | 96.56 / 100 | €23/month | Partial | ||
| 15 | Anthropic | 96.45 / 100 | Not available | Partial | ||
| 16 | Anthropic | 95.92 / 100 | Not available | Partial | ||
| 17 | 95.57 / 100 | €22/month | Partial | |||
| 18 | OpenAI | 95.41 / 100 | €23/month | Partial | ||
| 19 | Meta | 95.21 / 100 | Not available | Partial | ||
| 20 | 94.50 / 100 | €22/month | Partial | |||
| 21 | xAI | 94.25 / 100 | Free | Partial | ||
| 22 | Anthropic | 93.62 / 100 | Not available | Partial | ||
| 23 | Anthropic | 93.09 / 100 | Not available | Partial | ||
| 24 | Anthropic | 93.03 / 100 | Free | Partial | ||
| 25 | DeepSeek | 92.73 / 100 | Free | Partial | ||
| 26 | OpenAI | 92.72 / 100 | €23/month | Partial | ||
| 27 | OpenAI | 92.55 / 100 | €23/month | Partial | ||
| 28 | xAI | 92.54 / 100 | Free | Partial | ||
| 29 | OpenAI | 92.45 / 100 | €23/month | Partial | ||
| 30 | Anthropic | 92.38 / 100 | Not available | Partial | ||
| 31 | Anthropic | 92.37 / 100 | Not available | Partial | ||
| 32 | Alibaba | 92.36 / 100 | Not available | Partial | ||
| 33 | 92.20 / 100 | €22/month | Partial | |||
| 34 | OpenAI | 92.19 / 100 | €23/month | Partial | ||
| 35 | Anthropic | 92.02 / 100 | Not available | Partial | ||
| 36 | xAI | 92.01 / 100 | Free | Partial | ||
| 37 | xAI | 91.84 / 100 | Free | Partial | ||
| 38 | Anthropic | 91.68 / 100 | €23/month | Partial | ||
| 39 | Anthropic | 91.67 / 100 | €23/month | Partial | ||
| 40 | Baidu | 91.13 / 100 | Not available | Partial | ||
| 41 | xAI | 90.96 / 100 | Free | Partial | ||
| 42 | Mimo V2.5 Pro | Xiaomi | 90.95 / 100 | Not available | Partial | |
| 43 | DeepSeek | 90.78 / 100 | Free | Partial | ||
| 44 | Z.ai (zhipu ai) | 90.77 / 100 | Not available | Partial | ||
| 45 | OpenAI | 90.76 / 100 | €23/month | Partial | ||
| 46 | xAI | 90.75 / 100 | Free | Partial | ||
| 47 | Alibaba | 90.74 / 100 | Not available | Partial | ||
| 48 | OpenAI | 90.73 / 100 | €23/month | Partial | ||
| 49 | Anthropic | 90.66 / 100 | €23/month | Partial | ||
| 50 | DeepSeek | 90.40 / 100 | Free | Reliable |
Indicative score, weighted from public sources. Transparent methodology. See the AI ranking page for the full methodology and the other categories.
AI cost / performance ranking
Le Recul's Cost / performance ranking sets what a model is worth against what it actually costs in use. It draws on overall performance, price, speed, latency, stability and usable context — not on the headline price in isolation.
This is the most useful category when budget is a real constraint. Price gaps between models are far wider than quality gaps: on many everyday uses, a model two to ten times cheaper does an equivalent job. Paying for the top-ranked model only makes sense if your use genuinely requires what sets it apart.
The ranking does not tell you what to buy. It shows where the tipping point sits between what you pay and what you get, so the trade-off is made on figures rather than on an impression.
What this ranking is for
This ranking helps you avoid over-specifying: picking the most expensive model « just in case » is the most costly reflex and rarely the most justified.
It also helps spot the opposite case, where a cheap model ends up costing more in rework, latency or instability than a better-placed one.
How to read the ranking
- Overall quality, weighted as in the general ranking.
- Effective price in use, input and output.
- Speed and latency, which govern interactive use.
- Stability: consistency of the service and of the answers.
- Usable context that can actually be exploited, not the theoretical limit announced.
FAQ — Cost / performance
Which AI offers the best value for money?
It depends on volume and use. The ranking sets performance against cost so that the trade-off is explicit, rather than imposing a single choice.
Does this ranking use subscriptions or API prices?
This category reasons on the effective cost in use. For consumer assistants billed as a monthly subscription, the Chat ranking is the reference.
Is a cheaper model necessarily worse?
No, and that is the whole point of this category. On many everyday uses, the quality gap is far smaller than the price gap.
Why does usable context differ from the announced context?
Because answer quality often degrades before the technical limit. The context you can genuinely exploit is usually smaller than the advertised window.
How often is the ranking updated?
About every 25 hours, automatically. The update date appears at the top of the page.