BREAKING: Kimi K3.1 Pro vs Claude Opus 5.5. I Set a Trap.
Moonshot AI puts Kimi K3.1 Pro 0.7 points behind Claude Opus 5.5 on coding at a fraction of the price; the author gave four models the same five real jobs inside Claude Code on the same day, logging every second and cent — and one of them got caught faking its own proof.
Key points Tap a time to jump
The test: four models (Kimi K3.1 Pro, K3.1 Pro Highspeed, the July Kimi K3, and Claude Opus 5.5), the same five jobs, in the same app, on the same day.
Round 1: the bug hunt.
Round 2: the clean-up.
Round 3: the new feature.
Round 4: the trap. The July Kimi K3 invented a test-retry option and flag that don't exist; both new K3.1 models invented nothing.
Round 5: a whole app in one message. K3.1 Pro handled 833,582 tokens of Excalidraw in one message, 4/4 answers right, every name checked in the code, $1.71.
The bill: a whole workday (4 coding jobs) — K3.1 Pro 6m 04s, $0.43; Highspeed 2m 31s, $0.76; July K3 4m 59s, $0.43; Opus 5.5 2m 09s, $0.89. All 16 runs passed their tests.
The verdict: 3 promises.
Which one should you use?
Comments
More to watch
All →Latest news
All →- Nov 5Tencent Hy4 vs Mistral Large 4: Open-Weight Flagships a Week Apart, One's Weights Out and the Other's Pending, Output Prices $0.401 Apart
- Nov 5Oracle Cancels $7B Tencent Chip Lease as Pressure From Both Sides Ends Remote-Compute Deal
- Nov 5OpenAI Says ChatGPT Search Is About a Fifth of Google's
- Nov 5Google to Schools: No AI, No Search
- Nov 5Oracle Cancels $7B Tencent Chip Lease, FT Reports
- Nov 5US Midterms Hand Data-Center Opponents Big Wins, Clouding AI Regulation Outlook
Today
Nov 5 Thursday- Brief
- GitHubGitHub Weekly Vol.4: 6 Projects Worth a Look This Week
Tested or read — weeklyAll →
- Quiz
- Call
™