tweetindex

shirish @shiri_shh · 03 Sep 2026

The "single best LLM" does not exist. Martian evaluated 21 top models on 16 benchmarks and found that routing queries cuts error rates by 46% to 54% at the exact same cost. Benchmark leaderboards measure isolated models. Production systems need orchestration. If your https://t.co/2wGj1k3KhD
741Views
4Likes
0Reposts
1Replies
0Quotes
2Bookmarks

Is that a lot?

0.01×vs this author's median55 421 views is typical
11Percentile for this authorof 9 recent posts
0.22×vs 10K–100K median3 383 views is typical
1.68%Reachviews ÷ followers
0.68%Engagement rateof viewers reacted

Compare with the benchmark table →

Open on X →