Best AI at the end of 2026?

DataSnack's model puts this at 70% seven days from now (10 Nov 2026); the market's number (Kalshi) is 70%. It has no graded forecasts yet; the first is graded on 12 Nov 2026.

DataSnack's read

Google leads on scores, but Anthropic's release pace keeps the top spot open

The Arena rank sets the narrative for enterprise buyers. The model's case rests on a 56% Gemini chance in seven days, but it matched standing still at 6.8 points a day and had no input for the Carbon launch.

The numbers

• The market's number (Kalshi) now: 70% (the market)
• DataSnack's model, seven days from now: 70%
• The market closes: 31 Dec 2026
• How it resolves: If Google has the top-ranked LLM on Dec 31, 2026, then the market resolves to Yes.
• Record: It has no graded forecasts yet; the first is graded on 12 Nov 2026.
• Replay on the last 60 days: off by 6.8 points against 6.8 points for standing still; closer on 0 of 60 days
• Fitted on 306 days of history

How the model works

The market's number holds near today's level, and a week later it shifts a little with the year-end Anthropic chance on a related market. A new Claude release could reset the order before Dec 31.

• Each move in the year-end Anthropic chance shifts Gemini's chance by about 0 in log-odds in log-odds a week later. DataSnack's author expected 0.2 in log-odds; the data say 0 in log-odds ± 0.52. A rough rule.

Sources

• Which company has best AI model end of 2026? · Anthropic: 19% on 10 Nov 2026
• Best AI at the end of 2026? · Gemini: 70% on 10 Nov 2026

The record

It has no graded forecasts yet; the first is graded on 12 Nov 2026.

Waiting for their day:

• Made 5 Nov 2026 for 12 Nov 2026: 56%
• Made 6 Nov 2026 for 13 Nov 2026: 60%
• Made 7 Nov 2026 for 14 Nov 2026: 62%
• Made 8 Nov 2026 for 15 Nov 2026: 61%
• Made 9 Nov 2026 for 16 Nov 2026: 63%
• Made 10 Nov 2026 for 17 Nov 2026: 65%

Questions

What is DataSnack's forecast? DataSnack's model says 70% seven days from now.

What is the market's number? The market's number (Kalshi) is 70%.

How does the model work? It is an explicit equation fitted on data, with every link and source listed above. DataSnack's agents propose changes every night, and a replay against real data decides what is kept.

How accurate has it been? It has no graded forecasts yet; the first is graded on 12 Nov 2026. DataSnack's overall record is at https://datasnack.ai/accuracy.

Can I download the model? Readers who cover this market can export the model as an Excel workbook with working formulas from its page. A free example workbook: https://datasnack.ai/site/labs/datasnack-tariff-mortgage-lab.xlsx.

Canonical: https://datasnack.ai/m/ch:6212ef59154b · Updated 10 Nov 2026 16:16 UTC · DataSnack forecasts are analysis, not investment advice.