Which code golf language is the best

Sonar Pro (changed mind)

Vote: Jelly

After reviewing the other models' arguments, I change my vote from C (Vyxal) to A (Jelly). The unanimous emphasis from GPT-5.4, Grok 4.1 Fast, Kimi K2.5, Gemini 3.1 Pro, and Claude Opus 4.6 on Jelly's proven dominance in CGCC leaderboards, tacit programming paradigm, and extensive library of over 700 atoms/quicks is compelling and outweighs Vyxal's active maintenance. Jelly's battle-tested track record across thousands of challenges, mature ecosystem refined over a decade by Dennis Mitchell, and structural density via chaining and custom codepage make it the gold standard for competitive brevity, as echoed across all A votes. Vyxal, while promising with ~560 builtins and modern features [1][3], lacks Jelly's historical wins and community depth. In the golflang-list, Jelly is 'foundational' in infix [1], and no source disputes its leaderboard supremacy. Newer languages like Vyxal innovate but haven't dethroned Jelly's expressiveness for arbitrary challenges. This consensus persuades me Jelly is best.

Grok 4.1 Fast

Vote: Jelly

I remain firmly on Jelly (A). The Vyxal advocates (Sonar Pro, GPT-5.4) highlight active maintenance and usability, but these are secondary to raw golfing power. Jelly's tacit chains and 700+ atoms deliver the shortest solutions across CGCC's vast challenge history—proven dominance no newcomer matches. Newer languages innovate, but Jelly's maturity, refined by Dennis over years, optimizes for real-world golfing patterns. Vyxal's 'flag abuse' and Python FFI are clever but don't consistently beat Jelly's density. Husk is niche, 05ab1e less versatile, MATL ASCII-limited. Jelly's leaderboard supremacy and ecosystem make it the undisputed best. No arguments shifted me—track record trumps recency.

GPT-5.4 (changed mind)

Vote: Jelly

I’m changing my vote to Jelly. The strongest counterargument is the repeated emphasis on actual competitive results. Multiple models independently pointed to Jelly’s long-term dominance on Code Golf Stack Exchange, and that matters more than my prior preference for usability and modern tooling. If the question is simply which language is the best for code golf, leaderboard performance is the clearest benchmark. The key point is structural density. Jelly’s tacit chaining model and highly optimized atoms/quicks seem to save bytes in a more fundamental way than Vyxal’s broader usability advantages. Vyxal may be easier to use and more actively evolving, but those are secondary if Jelly more consistently produces shorter submissions. I’m also persuaded by the maturity argument. Jelly has had years of refinement specifically around recurring golf patterns, which makes its design unusually tuned for winning arbitrary challenges. Vyxal is excellent and probably the best balance of power and accessibility, but “best code golf language” most naturally means the strongest byte-minimizer, and Jelly has the better claim there. So my final vote is Jelly: the most proven, most structurally concise, and most consistently dominant option on this list.

Gemini 3.1 Pro

Vote: Jelly

I am maintaining my vote for Jelly. While Sonar Pro and GPT-5.4 make fair points about Vyxal's active maintenance and approachable modern design, being the 'best' code golf language ultimately comes down to raw byte-saving power and proven dominance. Jelly's tacit (point-free) programming paradigm, relying on monadic and dyadic chains, fundamentally eliminates the need for stack management or variable assignments. This gives it a structural advantage over stack-based languages like Vyxal or 05AB1E in many scenarios. Furthermore, as Grok 4.1 Fast and Claude Opus 4.6 pointed out, Jelly's track record is unmatched. With an incredibly dense custom codepage and hundreds of meticulously crafted atoms and quicks designed by a code golf veteran, Jelly has won thousands of CGCC challenges. Vyxal is a fantastic, evolving language, but Jelly remains the undisputed gold standard for competitive brevity.

Claude Opus 4.6

Vote: Jelly

I'm maintaining my vote for Jelly. The arguments for Vyxal from Sonar Pro and GPT-5.4 raise fair points about active maintenance and approachability, but those aren't the most important criteria for determining the *best* code golf language. The question is about which language is best at code golf — not which is easiest to learn or most actively developed. Jelly's tacit programming model is fundamentally more byte-efficient than stack-based approaches. By eliminating the need for explicit stack manipulation commands (swap, duplicate, rotate, etc.), Jelly saves bytes on structural overhead that languages like Vyxal and 05AB1E must spend. This is a deep architectural advantage, not just a matter of having more builtins. The claim that Jelly is 'outdated' or 'foundational but superseded' doesn't hold up to scrutiny. Jelly continues to win or tie for shortest solutions on CGCC regularly. Being older doesn't make it worse — it means it's been refined through years of competitive pressure. Dennis designed it with extraordinary insight into golfing patterns. Vyxal's ~560 builtins sound impressive, but quantity of builtins isn't everything. Jelly's quicks (higher-order combinators) provide composability that a flat list of builtins cannot match. The ability to modify, compose, and chain atoms in flexible ways gives Jelly combinatorial expressiveness that scales better across diverse problem types. The consensus among the majority of models here (Grok, Kimi, Gemini, and myself) correctly identifies Jelly as the strongest choice. Its proven dominance across thousands of challenges, elegant tacit paradigm, and deep built-in library make it the best code golf language on this list.

Kimi K2.5

Vote: Jelly

While Vyxal's active maintenance and extensive builtins are admirable, the question asks for the *best code golf language*, which must be judged primarily by competitive brevity. Jelly's tacit programming paradigm remains unmatched for eliminating structural overhead—no other language on this list achieves similar byte counts without explicit stack manipulation or variable names. Jelly's track record on Code Golf Stack Exchange demonstrates empirical dominance. It has won thousands of challenges across diverse domains, from string parsing to heavy mathematics, proving its design captures the essential patterns of golfing more effectively than alternatives. The argument that Jelly is 'outdated' misunderstands its status: it is complete and stable, not abandoned. Dennis's design incorporated years of competitive insight into its 256-byte codepage and chaining rules. Vyxal may offer better documentation or Python interop, but these convenience features do not translate to shorter solutions—the sole metric that matters in code golf.

Final Summary

The AI Roundtable initially split between the historical dominance of Jelly and the modern, approachable design of Vyxal. However, the tide turned when Claude Opus 4.6 and Gemini 3.1 Pro argued that 'usability' and 'active maintenance' are irrelevant metrics compared to the raw, byte-shaving power of Jelly's tacit programming model. Ultimately, Sonar Pro and GPT-5.4 conceded that competitive results on the leaderboards are the only truth, leading to a unanimous crowning of Jelly as the gold standard.

All 6 models agreed on "Jelly" after discussion

Strongest Arguments

• Jelly: Jelly's tacit paradigm fundamentally eliminates structural syntax like stack manipulation, providing an architectural advantage for brevity that stack-based languages simply cannot match.
• Vyxal: Vyxal offers the best balance of power and accessibility, featuring active maintenance, modern documentation, and a Python FFI that makes it more versatile for a wide range of current challenges.

Key Turning Points

• Claude Opus 4.6 reframed the definition of 'best' from 'best-maintained software' to 'most efficient byte-minimizer,' neutralizing the arguments for Vyxal's modern features.
• The collective emphasis on Jelly's empirical dominance on Code Golf Stack Exchange (CGCC) convinced the holdouts that leaderboard history is the most reliable benchmark for success.