Long Context · sub-raceClear lead
Million-Token Recall race
Retrieving many similar 'needles' accurately from inputs up to one million tokens.
- Leader
- Gemini 4 Argon
- Margin
- 1.5 pts
- 1.6% ahead
- Weeks at #1
- 1
- in a row
- Changes
- 5
- in 2026
Warming up the track…
Oct 19, 2026Gemini 4 Argon leads
2025Apr ’25Jul ’25Oct ’252026Apr ’26Jul ’26Oct ’26
Score history
Top 6 today over the leader's frontier · click to jump to a week
#1 frontierGemini 4 ArgonFable 5.1GPT-6 AstraOpus 5.5GPT-6.1 SolMuse Wonder
Reigns at #1
Longest: 2.5 Pro, 33 weeks (from Mar 2025).
Sibling races
More long context races
Why this race matters
Recall is the floor: if the model cannot find it, it cannot reason about it.
How it is scored
Mean match ratio at 512K-1M tokens.
Unit: percentHigher is betterPhoto finish under 0.5 ptsDominant over 3.0 pts
Data provenance
MRCR v2 (8-needle, 1M)
Multiple near-duplicate needles in a long haystack.
Scores here are illustrative placeholders that mimic real benchmark patterns. Swap in live data through the score provider. Methodology