Claude Haiku 5.5 · Coding
Claude Haiku 5.5 for coding: our test runs
Claude Haiku 5.5 ran our 5 coding prompts in llmwise, through the same pipeline your messages take. Here's every reply as it came, whether it passed, what it cost and what it counts as on Pro, next to the other models.
Based on 5 of our test runs on , through OpenRouter with the app's own prompt and settings.
Short answer
In our test runs on November 18, 2026, Claude Haiku 5.5 passed 5 of 5 of our coding prompts (both hard ones): tied 1st of 27 models. Each reply counted as 1 message on Pro, where it has 60 a day, and cost $0.0004 to run on average.
Prompt by prompt
Each of the 5 coding prompts, and how Claude Haiku 5.5's reply did.
| Prompt | Result | Cost | Time |
|---|---|---|---|
| Turn a title into a URL slugEveryday | Passed: All 7 tests passed. | $0.0002 | 1.3 s |
| Parse a duration like “1h 30m”Everyday | Passed: All 10 tests passed. | $0.0001 | 1.2 s |
| Merge overlapping intervalsEveryday | Passed: All 7 tests passed. | $0.0001 | 1.1 s |
| Evaluate an arithmetic expression, no evalHard | Passed: All 15 tests passed. | $0.0013 | 9.3 s |
| Parse CSV with quoted fieldsHard | Passed: All 8 tests passed. | $0.0003 | 2.2 s |
Every coding reply from Claude Haiku 5.5
Turn a title into a URL slug
An everyday prompt: “Turn a title into a URL slug”, in full.
Passed: All 7 tests passed.
```javascript function slugify(title) { return String(title) .normalize("NFD") .replace(/[\u0300-\u036f]/g, "") .toLowerCase() .replace(/[^a-z0-9]+/g, "-") .replace(/^-+|-+$/g, ""); } ```926 tokens in, 112 out · 1.3 s · $0.0002 · 1 message on Pro · answered by anthropic/claude-haiku-5.5 via Anthropic ·
Parse a duration like “1h 30m”
An everyday prompt: “Parse a duration like “1h 30m””, in full.
Passed: All 10 tests passed.
```javascript function parseDuration(text) { if (typeof text !== "string") return null; const s = text.trim(); const match = /^(?:(\d+)\s*h)?\s*(?:(\d+)\s*m)?\s*(?:(\d+)\s*s)?$/.exec(s); if (!match || s === "") return null; const [, h, m, sec] = match; if (h === undefined && m === undefined && sec === undefined) return null; return (Number(h ?? 0) * 3600) + (Number(m ?? 0) * 60) + Number(sec ?? 0); } ```881 tokens in, 212 out · 1.2 s · $0.0001 · 1 message on Pro · answered by anthropic/claude-haiku-5.5 via Anthropic ·
Merge overlapping intervals
An everyday prompt: “Merge overlapping intervals”, in full.
Passed: All 7 tests passed.
```javascript function mergeIntervals(intervals) { const sorted = intervals .map(([start, end]) => [start, end]) .sort((a, b) => a[0] - b[0] || a[1] - b[1]); const result = []; for (const [start, end] of sorted) { const last = result[result.length - 1]; if (last && start <= last[1]) { last[1] = Math.max(last[1], end); } else { result.push([start, end]); } } return result; } ```876 tokens in, 180 out · 1.1 s · $0.0001 · 1 message on Pro · answered by anthropic/claude-haiku-5.5 via Anthropic ·
Evaluate an arithmetic expression, no eval
A hard prompt: “Evaluate an arithmetic expression, no eval”, in full.
Passed: All 15 tests passed.
```javascript function evaluate(expression) { const src = String(expression); // Tokenize: numbers, operators, parentheses. Spaces are skipped. const tokens = []; for (let i = 0; i < src.length; ) { const c = src[i]; if (/\s/.test(c)) { i++; continue; } const m = /^(\d+(?:\.\d+)?|\.\d+)/.exec(src.slice(i)); if (m) { tokens.push({ type: 'num', value: Number(m[1]) }); i += m[1].length; continue; } if ('+-*/^()'.includes(c)) { tokens.push({ type: c }); i++; continue; } throw new Error(`Unexpected character "${c}" at position ${i}`); } // Recursive descent parser. // expr := term (('+' | '-') term)* // term := unary (('*' | '/') unary)* // unary := '-' unary | power // power := primary ('^' unary)? (right-associative, exponent may be signed) // primary := number | '(' expr ')' let p = 0; const peek = () => (p < tokens.length ? tokens[p].type : null); const expect = (type) => { if (peek() !== type) { throw new Error(`Expected "${type}" at token ${p + 1}`); } p++; }; function parseExpr() { let v = parseTerm(); while (peek() === '+' || peek() === '-') { const op = tokens[p++].type; const r = parseTerm(); v = op === '+' ? v + r : v - r; } return v; } function parseTerm() { let v = parseUnary(); while (peek() === '*' || peek() === '/') { const op = tokens[p++].type; const r = parseUnary(); v = op === '*' ? v * r : v / r; } return v; } function parseUnary() { if (peek() === '-') { p++; return -parseUnary(); } return parsePower(); } function parsePower() { const base = parsePrimary(); if (peek() === '^') { p++; return base ** parseUnary(); } return base; } function parsePrimary() { if (peek() === 'num') { return tokens[p++].value; } if (peek() === '(') { p++; const v = parseExpr(); expect(')'); return v; } if (peek() === null) { throw new Error('Unexpected end of expression'); } throw new Error(`Unexpected token "${tokens[p].type}" at token ${p + 1}`); } if (tokens.length === 0) { throw new Error('Empty expression'); } const result = parseExpr(); if (p !== tokens.length) { throw new Error(`Unexpected token "${tokens[p].type}" at token ${p + 1}`); } return result; } ```1,017 tokens in, 2,412 out (1,349 of them reasoning) · 9.3 s · $0.0013 · 1 message on Pro · answered by anthropic/claude-haiku-5.5 via Anthropic ·
Parse CSV with quoted fields
A hard prompt: “Parse CSV with quoted fields”, in full.
Passed: All 8 tests passed.
```javascript function parseCsv(text) { const rows = []; let row = []; let field = ""; let inQuotes = false; let fieldStarted = false; let i = 0; const n = text.length; if (n === 0) return []; while (i < n) { const ch = text[i]; if (inQuotes) { if (ch === '"') { if (text[i + 1] === '"') { field += '"'; i += 2; continue; } inQuotes = false; i++; continue; } field += ch; i++; continue; } if (ch === '"' && !fieldStarted) { inQuotes = true; fieldStarted = true; i++; } else if (ch === ",") { row.push(field); field = ""; fieldStarted = false; i++; } else if (ch === "\r" || ch === "\n") { row.push(field); rows.push(row); row = []; field = ""; fieldStarted = false; if (ch === "\r" && text[i + 1] === "\n") i += 2; else i++; // A trailing line break ends the last row without starting a new one. if (i >= n) return rows; } else { field += ch; fieldStarted = true; i++; } } // Flush the final row (text did not end with a line break). row.push(field); rows.push(row); return rows; } ```924 tokens in, 513 out · 2.2 s · $0.0003 · 1 message on Pro · answered by anthropic/claude-haiku-5.5 via Anthropic ·
How it compares on coding
Claude Haiku 5.5 beside the job's picks among every model. 25 models share the hard-problems pick, Claude Haiku 5.5 among them.
| Model | Passed | On Pro | Cost per reply |
|---|---|---|---|
| Gemini 3.8 FlashOur pick: hard problems (shared), best value | 5 of 5 | 250 a month on Pro | $0.0011 |
| GPT-6 LunaOur pick: hard problems (shared), everyday | 5 of 5 | 60 a day on Pro | $0.0001 |
| Claude Haiku 5.5Our pick: hard problems (shared) | 5 of 5 | 60 a day on Pro | $0.0004 |
Every limit is published. Paid plans also have a monthly fair-use limit on AI cost: Plus $3.75, Pro $7.50, Max $20, Ultra $42, Studio $85. Using every message on your plan at typical sizes stays under it; very large messages and heavy research use it faster. Every limit, explained.
Claude Haiku 5.5 next to each model on coding
Claude Haiku 5.5 against each model on the same 5 coding prompts. One run each, through OpenRouter: a wait depends on the provider and the load that day, so a lead under 10% counts as close.
Against Claude Fable 5.2: Claude Haiku 5.5 answered 3.2× sooner (1.3 s to 4.1 s), cost 62.7× less, and passed 5 of 5 to its 5.
Against Claude Fable 5.1: Claude Haiku 5.5 answered 3.5× sooner (1.3 s to 4.5 s), cost 84.0× less, and passed 5 of 5 to its 5.
Against Claude Opus 5.5: Claude Haiku 5.5 answered 4.5× sooner (1.3 s to 5.7 s), cost 45.2× less, and passed 5 of 5 to its 5.
Against Claude Sonnet 5.5: Claude Haiku 5.5 answered 1.9× sooner (1.3 s to 2.5 s), cost 12.8× less, and passed 5 of 5 to its 5.
Against Claude Sonnet 5: Claude Haiku 5.5 answered 2.1× sooner (1.3 s to 2.6 s), cost 25.4× less, and passed 5 of 5 to its 5.
Against Claude Haiku 4.5: Claude Haiku 5.5 answered 1.7× sooner (1.3 s to 2.2 s), cost 7.5× less, and passed 5 of 5 to its 3. Only Claude Haiku 5.5 passed Evaluate an arithmetic expression, no eval; Parse CSV with quoted fields.
Against GPT-6.2 Astra: Claude Haiku 5.5 answered 2.7× sooner (1.3 s to 3.5 s), cost 33.3× less, and passed 5 of 5 to its 5.
Against GPT-6 Astra: Claude Haiku 5.5 answered 4.5× sooner (1.3 s to 5.8 s), cost 27.3× less, and passed 5 of 5 to its 5.
Against GPT-6.2 Sol: Claude Haiku 5.5 answered 2.9× sooner (1.3 s to 3.8 s), cost 9.4× less, and passed 5 of 5 to its 5.
Against GPT-6.1 Sol: Claude Haiku 5.5 answered 6.1× sooner (1.3 s to 7.9 s), cost 10.3× less, and passed 5 of 5 to its 5.
Against GPT-6 Sol: Claude Haiku 5.5 answered 3.7× sooner (1.3 s to 4.8 s), cost 7.1× less, and passed 5 of 5 to its 5.
Against GPT-6.1 Terra: Claude Haiku 5.5 answered 5.3× sooner (1.3 s to 6.8 s), cost 3.4× less, and passed 5 of 5 to its 5.
Against GPT-6 Luna: Claude Haiku 5.5 answered 5.7× sooner (1.3 s to 7.3 s), cost 2.8× more, and passed 5 of 5 to its 5.
Against Gemini 4 Carbon: Claude Haiku 5.5 answered 28.3× sooner (1.3 s to 36.4 s), cost 35.4× less, and passed 5 of 5 to its 5.
Against Gemini 4 Argon: Claude Haiku 5.5 answered 5.9× sooner (1.3 s to 7.6 s), cost 6.5× less, and passed 5 of 5 to its 5.
Against Gemini 4 Hydrogen: Claude Haiku 5.5 answered 5.3× sooner (1.3 s to 6.9 s), cost 15.4× less, and passed 5 of 5 to its 5.
Against Gemini 3.1 Pro (preview): Claude Haiku 5.5 answered 12.2× sooner (1.3 s to 15.7 s), cost 33.0× less, and passed 5 of 5 to its 5.
Against Gemini 3.8 Flash: Claude Haiku 5.5 answered 2.5× sooner (1.3 s to 3.3 s), cost 2.8× less, and passed 5 of 5 to its 5.
Against DeepSeek V4.1 Flash: Claude Haiku 5.5 answered 2.7× sooner (1.3 s to 3.5 s), cost 3.7× less, and passed 5 of 5 to its 5.
Against DeepSeek V4 Pro: Claude Haiku 5.5 answered 8.8× sooner (1.3 s to 11.3 s), cost 29.6× less, and passed 5 of 5 to its 5.
Against Grok 4.7: Claude Haiku 5.5 answered 24.2× sooner (1.3 s to 31.2 s), cost 68.3× less, and passed 5 of 5 to its 5.
Against Kimi K3.1 Pro: Claude Haiku 5.5 answered 6.1× sooner (1.3 s to 7.8 s), cost 18.7× less, and passed 5 of 5 to its 5.
Against Kimi K3: Claude Haiku 5.5 answered 4.8× sooner (1.3 s to 6.2 s), cost 18.6× less, and passed 5 of 5 to its 5.
Against GLM 5.3: Claude Haiku 5.5 answered 1.2× sooner (1.3 s to 1.5 s), cost 5.7× less, and passed 5 of 5 to its 3. Only Claude Haiku 5.5 passed Parse a duration like “1h 30m”; Parse CSV with quoted fields.
Against GLM 5.3 Flash: Claude Haiku 5.5 answered 1.2× later (1.3 s to 1.1 s), cost 1.5× more, and passed 5 of 5 to its 5.
Against Mistral Large 4: Claude Haiku 5.5 answered 36.5× sooner (1.3 s to 46.9 s), cost 18.4× less, and passed 5 of 5 to its 5.
How these runs were done
Tests. Automatic. The function runs against the prompt's tests in a separate Node.js process with a time limit and no file, network or child-process access; it passes when every test passes.
Claude Haiku 5.5, and coding, elsewhere
- Claude Haiku 5.5: price, limits and messages on every plan
- The best AI for coding: our picks among every model
- Claude for coding
- Claude Haiku 5.5 for writing: our test runs
- Claude Haiku 5.5 for math: our test runs
- Claude Haiku 5.5 for data analysis: our test runs
- Claude Haiku 5.5 vs GPT-6 Luna
- Claude Haiku 5.5 vs Claude Haiku 4.5
- Claude Sonnet 5.5 vs Claude Haiku 5.5
- Claude Haiku 5.5 vs GLM 5.3 Flash
- Claude Haiku 5.5 vs DeepSeek V4.1 Flash
- Analyze a CSV with code: the prompt, and its price
- AI prompts for coding, checked by running the code
- Our test runs: 50 prompts on every model
Questions
Is Claude Haiku 5.5 good for coding?
In our test runs it passed 5 of 5 coding prompts, tied 1st of the 27 models in llmwise. Every reply is on this page, so you can judge them yourself.
How many of my messages does a coding reply from Claude Haiku 5.5 use?
1 message each on Pro, where it has 60 a day on Pro. The price of a message is fixed and shown before you send it, however long the reply.
How were these runs done?
The same way for every model: each prompt sent through llmwise's own pipeline, each reply checked the same way. The methods page has every prompt and how each is scored.
Claude, GPT, Gemini, DeepSeek, Grok, Kimi, GLM, and Mistral, in one chat.
See what a message costs before you send it. Every plan is month to month: cancel in two clicks. Sign in with an email link, no password.