OpenDotsStart here

CODING COMPARISON · OCTOBER 5, 2026

GPT-6.1 Sol vs Opus 5.5, in four panels

In one reported Go interpreter test, Sol’s code ran faster on a small calculation. Opus’s code handled infinite recursion without crashing the process. The finish line was only part of the test.

Four-panel comic: both implementations pass 206 cases; Sol runs fib(27) faster; infinite recursion crashes Sol’s process while Opus returns an error.
Original comic adapted from Thomas Wiegold’s reported test. Dialogue and scenes are dramatized.
Sol and Opus hold certificates showing 206 out of 206 cases passed.
01. Both passed the regular tests.
The coworkers race with computers; the scoreboard compares interpreter runtimes, not model response times.
02. Sol’s generated interpreter ran fib(27) in 31 ms; Opus’s took 45 ms.
Sol continues toward a cliff while Opus stops at a recursion limit.
03. One more edge case: infinite recursion.
Sol’s computer crashes; Opus returns an error. The cliff and dialogue are comic metaphors.
04. “I asked who finishes first. Not who disappears first.”

Author-reported results, not a test performed by OpenDots. Company marks identify the characters; this is an independent editorial comparison.

Open the full-size comicRead Thomas Wiegold’s original review ↗

The results behind the joke

Generated Go interpreters in the author’s test
CheckGPT-6.1 SolOpus 5.5
Handwritten conformance cases206 / 206 passed206 / 206 passed
fib(27) execution time31 ms45 ms
Unbounded recursionProcess crashedError returned

These timings describe the generated programs, not how quickly the models answered. The reviewer also reports that Opus’s depth limit rejected some extremely deep, valid programs. This is one task, not a universal ranking.

What to check in your own code

Use known answers and failure cases alongside the happy path. Check whether generated code stops safely when inputs exceed its limits, and whether those limits reject work you actually need.

Comic transcript

  1. Both coworkers hold certificates: “206 / 206. All tests passed!”
  2. Sol says, “Mine runs faster.” The fib(27) scoreboard shows Sol 31 ms and Opus 45 ms. These are generated interpreter runtimes.
  3. The client says, “One more edge case.” The sign reads “Infinite recursion”; Opus holds a stop sign.
  4. Sol’s generated interpreter crashes. Opus returns a recursion-limit error. The client says, “I asked who finishes first. Not who disappears first.”

Source and illustration

Based on Thomas Wiegold’s GPT-6.1 Sol review, published October 5, 2026. We have not reproduced the benchmark. The race, cliff, characters and dialogue are editorial inventions.

AI-generated illustration using the Orange Line Illustration skill. OpenAI and Anthropic marks identify the models’ companies. OpenDots is not affiliated with either company.