Best AI · Data analysis
The best AI for data analysis, from our test runs
We gave every model the same small tables of orders and ad spend and asked questions with one right answer: a top region, an average, a change, a filtered median and a correlation. GPT-6 Luna, Claude Haiku 5.5, DeepSeek V4.1 Flash, GLM 5.3, DeepSeek V4 Pro, Mistral Large 4, Grok 4.7, GPT-6.1 Sol, GPT-6 Sol, Claude Sonnet 5.5, Kimi K3, Gemini 3.1 Pro, Claude Opus 5.5, GPT-6 Astra, and Claude Fable 5.1 each passed 5 of 5, the most; GPT-6 Luna is first on the tie-break (the smaller message, then the lower cost).
Test runs and prices checked . Updated .
Short answer
GPT-6 Luna, Claude Haiku 5.5, DeepSeek V4.1 Flash, GLM 5.3, DeepSeek V4 Pro, Mistral Large 4, Grok 4.7, GPT-6.1 Sol, GPT-6 Sol, Claude Sonnet 5.5, Kimi K3, Gemini 3.1 Pro, Claude Opus 5.5, GPT-6 Astra, and Claude Fable 5.1 each passed 5 of 5, the most; GPT-6 Luna is first on the tie-break (the smaller message, then the lower cost). This is our own test, run on October 9, 2026: every prompt, reply and score is published.
Ranked by our test runs
This is our own test: the same prompts sent to every model through llmwise, each reply checked the same way, with every prompt, reply and score published. The data prompts give each model a small table as CSV and check its final number; they don't test charts, spreadsheet formulas or large files.
| # | Model | Passed | Hard ones | Free trial | On Pro | Cost per reply |
|---|---|---|---|---|---|---|
| 1 | GPT-6 LunaOpenAI | 5 of 5 | 2 of 2 | Yes | 60 a day | $0.0002 |
| 2 | Claude Haiku 5.5Anthropic | 5 of 5 | 2 of 2 | Yes | 60 a day | $0.0004 |
| 3 | DeepSeek V4.1 FlashDeepSeek | 5 of 5 | 2 of 2 | Yes | 60 a day | $0.0008 |
| 4 | GLM 5.3Z.ai | 5 of 5 | 2 of 2 | Yes | 250 a month | $0.0019 |
| 5 | DeepSeek V4 ProDeepSeek | 5 of 5 | 2 of 2 | Yes | 250 a month | $0.0029 |
| 6 | Mistral Large 4Mistral | 5 of 5 | 2 of 2 | Yes | 250 a month | $0.0037 |
| 7 | Grok 4.7xAI | 5 of 5 | 2 of 2 | Yes | 250 a month | $0.0111 |
| 8 | GPT-6.1 SolOpenAI | 5 of 5 | 2 of 2 | Yes | 125 a month | $0.0016 |
| 9 | GPT-6 SolOpenAI | 5 of 5 | 2 of 2 | Yes | 125 a month | $0.0034 |
| 10 | Claude Sonnet 5.5Anthropic | 5 of 5 | 2 of 2 | Yes | 125 a month | $0.0047 |
| 11 | Kimi K3Moonshot | 5 of 5 | 2 of 2 | Yes | 125 a month | $0.0068 |
| 12 | Gemini 3.1 Pro (preview)Google | 5 of 5 | 2 of 2 | Yes | 125 a month | $0.0155 |
| 13 | Claude Opus 5.5Anthropic | 5 of 5 | 2 of 2 | 1 message | 62 a month | $0.0131 |
| 14 | GPT-6 AstraOpenAI | 5 of 5 | 2 of 2 | No | 31 a month | $0.0154 |
| 15 | Claude Fable 5.1Anthropic | 5 of 5 | 2 of 2 | No | 31 a month | $0.0283 |
| 16 | GLM 5.3 FlashZ.ai | 4 of 5 | 1 of 2 | Yes | 60 a day | $0.0004 |
| 17 | Gemini 3.8 FlashGoogle | 4 of 5 | 1 of 2 | Yes | 250 a month | $0.0028 |
| 18 | Claude Sonnet 5Anthropic | 4 of 5 | 1 of 2 | Yes | 125 a month | $0.0069 |
| 19 | Claude Haiku 4.5Anthropic | 3 of 5 | 1 of 2 | Yes | 250 a month | $0.0025 |
The prompts, every reply and how each was scored: our test runs.
The best pick at each price
The same results, by what a message counts as on Pro: the best model at each size of message, then the rest of that size.
60 a day on Pro (everyday models)
GPT-6 Luna, 5 of 5 passed; then Claude Haiku 5.5 5 of 5, DeepSeek V4.1 Flash 5 of 5, and GLM 5.3 Flash 4 of 5.
250 a month on Pro
GLM 5.3, 5 of 5 passed; then DeepSeek V4 Pro 5 of 5, Mistral Large 4 5 of 5, Grok 4.7 5 of 5, Gemini 3.8 Flash 4 of 5, and Claude Haiku 4.5 3 of 5.
125 a month on Pro
GPT-6.1 Sol, 5 of 5 passed; then GPT-6 Sol 5 of 5, Claude Sonnet 5.5 5 of 5, Kimi K3 5 of 5, Gemini 3.1 Pro 5 of 5, and Claude Sonnet 5 4 of 5.
62 a month on Pro
Claude Opus 5.5, 5 of 5 passed.
31 a month on Pro
GPT-6 Astra, 5 of 5 passed; then Claude Fable 5.1 5 of 5.
Every limit is published. Paid plans also have a monthly fair-use limit on AI cost: 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.
One prompt, side by side
An everyday data analysis prompt, “Average order value in August”, where the models didn't all pass.
The prompt: Average order value in August (everyday)
Here are some orders as CSV. Revenue is quantity times price. What was the average revenue per order in August 2026, to 2 decimal places? order_id,date,region,product,quantity,price 1001,2026-07-03,West,Lamp,2,45.00 1002,2026-07-05,East,Chair,1,120.00 1003,2026-07-09,West,Desk,1,310.00 1004,2026-07-14,North,Lamp,4,45.00 1005,2026-07-21,East,Lamp,3,45.00 1006,2026-07-28,North,Chair,2,120.00 1007,2026-08-02,West,Chair,2,120.00 1008,2026-08-06,East,Desk,1,310.00 1009,2026-08-11,North,Desk,2,310.00 1010,2026-08-15,West,Lamp,5,45.00 1011,2026-08-19,East,Chair,3,120.00 1012,2026-08-27,North,Lamp,1,45.00 Show your working briefly, then end with a line that says "Final answer: " followed by the answer alone.
GPT-6 Luna
Passed: Final answer 300.00: right.
August revenue: $240 + $310 + $620 + $225 + $360 + $45 = $1,800. There were 6 orders, so the average is $1,800 ÷ 6 = $300.00. Final answer: $300.00
647 tokens in, 131 out (67 of them reasoning) · 2.6 s · $0.0001 · 1 message on Pro · answered by openai/gpt-6-luna via OpenAI ·
Claude Sonnet 5.5
Passed: Final answer 300.00: right.
August 2026 orders (1007–1012): - 1007: 2 × 120 = 240 - 1008: 1 × 310 = 310 - 1009: 2 × 310 = 620 - 1010: 5 × 45 = 225 - 1011: 3 × 120 = 360 - 1012: 1 × 45 = 45 Total = 240 + 310 + 620 + 225 + 360 + 45 = 1800 Average = 1800 / 6 = 300.00 Final answer: 300.00
825 tokens in, 170 out · 1.7 s · $0.0037 · 1 message on Pro · answered by anthropic/claude-sonnet-5.5 via Anthropic ·
Claude Haiku 4.5
Failed: Final answer 275.00; expected 300.
I'll find the average revenue per order for August 2026. **August 2026 orders:** | order_id | date | quantity | price | revenue | |----------|------|----------|-------|---------| | 1007 | 2026-08-02 | 2 | 45.00 | 90.00 | | 1008 | 2026-08-06 | 1 | 310.00 | 310.00 | | 1009 | 2026-08-11 | 2 | 310.00 | 620.00 | | 1010 | 2026-08-15 | 5 | 45.00 | 225.00 | | 1011 | 2026-08-19 | 3 | 120.00 | 360.00 | | 1012 | 2026-08-27 | 1 | 45.00 | 45.00 | **Calculation:** - Total revenue: 90 + 310 + 620 + 225 + 360 + 45 = 1,650.00 - Number of orders: 6 - Average revenue per order: 1,650.00 ÷ 6 = 275.00 Final answer: 275.00
683 tokens in, 326 out · 2.1 s · $0.0023 · 1 message on Pro · answered by anthropic/claude-haiku-4.5 via Amazon Bedrock ·
How we ranked them
Ranked by how many prompts each model passed, then how many of the hard ones, then by the smaller message (the everyday models first), then by the lower cost per reply. No ranking is chosen by hand.
- Our test runs: 50 prompts on every model
- The AI model leaderboard, job by job
- Best AI assistants, tested (October 2026)
- The best AI for research, from our test runs
- The best AI for students, from our test runs
- The best free AI, from our test runs
- The best AI model right now: all 19, ranked by our tests
- The best AI for business, from our test runs
- AI chat with every top model in one place
Questions
What is the best AI for data analysis?
In our own test runs on October 2026: GPT-6 Luna, Claude Haiku 5.5, DeepSeek V4.1 Flash, GLM 5.3, DeepSeek V4 Pro, Mistral Large 4, Grok 4.7, GPT-6.1 Sol, GPT-6 Sol, Claude Sonnet 5.5, Kimi K3, Gemini 3.1 Pro, Claude Opus 5.5, GPT-6 Astra, and Claude Fable 5.1 each passed 5 of 5, the most; GPT-6 Luna is first on the tie-break (the smaller message, then the lower cost); of the everyday models, GPT-6 Luna did best (5 of 5). Each answer is checked against the right number, so a pass means the arithmetic was right.
Is this the best AI for Excel?
Partly. These prompts test working out answers from a table (totals, averages, a filtered median, a correlation), which is the heart of spreadsheet analysis. They don't test writing Excel formulas or macros, so we don't rank models for that.
Can the AI run code on my data?
On paid plans the model can run code on a file you attach and show the code it ran. These runs don't use it: every model got the table in the message and had to work it out.
Claude, GPT, Gemini, DeepSeek, Grok, Kimi, GLM, and Mistral, in one chat.
See what a message costs before you send it. Free is 5 messages to try; sign in with an email link, no password or card.