iCreatAI News
Skip to content
Trending storyWatching

Frisson Labs releases language-controllable Melee model Reyn 26C

1 reports1 sources1 days agoUpdated

Understand this story

AI overview

Frisson Labs has released Reyn 26C, a language-controllable version of its Faynt model that extends language instruction coverage to all 26 characters in Super Smash Bros. Melee and adds 2v2 team battle support for the first time. According to figures published by Frisson Labs, the model achieved an 81.8% pass rate on instruction-following tests, an 84.4% win rate against two Faynt 2v2 live matches, and reduced damage to teammates by 71%–74% through in-game action masking. The model also comes with a voice dialogue pipeline that can convert speech into instructions and provide voice commentary on matches. These capabilities still have clear limitations: instruction support across the 26 characters is uneven, and special-move instructions for characters such as Mewtwo and Zelda remain unreliable; the experimental 2v2 branch currently uses only Fox and is independent of the currently released Reyn 26C.

AI overview based on source reports · Updated 1 days ago

Coverage timeline

Follow the coverage to explore different perspectives.

Nov 17
  1. Hacker News AI
    We got AI to Wombo Combo – collaborative agents in Melee

    The team behind Faynt introduces Reyn 26C, a language-steerable Super Smash Bros. Melee model built by adding labelled-replay imitation and RL to a frozen gameplay policy, then fine-tuning a text router to cover all 26 characters; it met a command-following target in 81.8% of test cases. A separate experimental Fox 2v2 branch, trained with full-policy multi-agent RL, won 84.4% of 160 real games against two Faynts, and a play-time action mask cut teammate damage by 71–74%. A bonus 2v1 combo model produced repeatable routines against an idle Bowser but far weaker results against opponents that could act.

Activity for this story

Current comparable activity
5
Peak comparable activity
10Nov 17 08:00
Change in the past 24 hours
-51%

Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.