7 ms·
I’ve tested this model on four of my benchmarks: https://github.com/lechmazur/buyout_game https://github.com/lechmazur/buyout_game 10th out 36. https://github
by zone411 4mo ago
I’ve tested this model on four of my benchmarks:
https://github.com/lechmazur/buyout_game https://github.com/lechmazur/buyout_game 10th out 36.
https://github.com/lechmazur/pact/ https://github.com/lechmazur/pact/ 14th out 25.
https://github.com/lechmazur/nyt-connections/ https://github.com/lechmazur/nyt-connections/ 60th out 81.
https://github.com/lechmazur/debate https://github.com/lechmazur/debate 16th out of 29.
- CamperBob2 4mo agoWould be interesting to see the 27B dense Qwen 3.6 model thrown into the mix.
- baxtr 4mo agoGood stuff! Is there a reason you change the leaderboard graphs for the third and fourth one? Also: would be great to have an overview page with a summary over all test, like a total score or similar.
- Sandworm5639 4mo agooh, I love the connections benchmark. Just curious, can you share what are those hardest puzzles that even the top models can't crack? sometimes when I find the puzzle absolutely undecipherable I like to ask LLMs to solve it, and I haven't seen them fail yet.