Compare / Mistral Large 4 vs GPT-6 Luna
Mistral Large 4 vs GPT-6 Luna
Which is cheaper, which is more capable, and which does better at each kind of work, from 89 results on 6 sources that measured both models.
Of the 89 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.
Results as of 19 November 2026 (catalogue release 2026-11-19-3608133e8331). See also Mistral AI vs OpenAI, every model from both labs.
Mistral Large 4 or GPT-6 Luna? The short answer
Pick GPT-6 Luna: it scores higher (Spring Prompt overall 38 against 24) and costs less.
- Spring Prompt overall: Mistral Large 4 24, GPT-6 Luna 38 (out of 100; how the score works).
- GPT-6 Luna is cheaper to run: $0.000 against $0.003 per product listing on CatalogBench (6.9 times less).
- They score about the same on the Artificial Analysis Intelligence Index (38.4 and 38.1).
- GPT-6 Luna writes faster: 127 tokens a second against 60.
- Mistral Large 4 is newer: released 6 Oct 2026, against 22 Sep 2026 for GPT-6 Luna.
- On our own benchmarks, on the results both models have: GPT-6 Luna does better for product listings, decks from an analysis, user surveys and marketing planning.
- On third-party benchmarks, on the results both models have: Mistral Large 4 does better for agents and tool use.
Head to head
Every fact side by side
| Mistral Large 4 | GPT-6 Luna | |
|---|---|---|
| Price per million tokens | $1.36 in · $4.18 out | $0.10 in · $0.50 out |
| Blended (3 in : 1 out) | $2.06 | $0.20 |
| Cheapest host, blended | $1.05 (Mistral) | $0.20 (Amazon Bedrock) |
| Cost per product listing (CatalogBench) | $0.003 | $0.000 |
| Cached input, per million | – | $0.010 cached input |
| Intelligence Index | 38.4 | 38.1 |
| Spring Prompt overall | 24 | 38 |
| Output speed, tokens a second | 60 | 127 |
| Context window | 1,048,576 tokens | 1,050,000 tokens |
| Released | 6 Oct 2026 (newer) | 22 Sep 2026 |
| Weights | Open | Closed |
| Developer based in | France | the US |
Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.
Which is better for what
Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.
Their slides, side by side
The same brief for both: Manchester store investment review for Hearthside Coffee, an invented company. First slide of each deck from DeckBench.
From our research
- AI Checkpoint: Sol’s new fast lane 9 Oct 2026
- Mistral Large 4 Is Fast, Cheap and Europe's Best Model. On Business Work, It Still Makes Things Up. 7 Oct 2026
- We Told 11 AI Models What Happened in 2026. Most of Them Thought We Were Making It Up. 12 Oct 2026
- Best AI for Product Descriptions: Accuracy, Cost and Real Examples 20 Oct 2026
Every shared result
Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.
| Measure | Mistral Large 4 | GPT-6 Luna |
|---|---|---|
| BulletBench measured by Spring Prompt · 3 results | ||
| Cost per game · Blitz 3+2US dollars, lower is better 2 · 1 of 23 measured | $0.0082 | $0.0072 |
| Games lost on time · Blitz 3+2% of games, lower is better 1 · 15 of 23 measured | 0.0% | 50.0% |
| Invalid moves · Blitz 3+2% of moves, lower is better 22 · 1 of 23 measured | 0.6% | 0.0% |
| CatalogBench measured by Spring Prompt · 3 results | ||
| Cost per productUS dollars, lower is better 8 · 2 of 36 measured | $0.0033 | $0.0005 |
| Reliably publish-ready% of products, higher is better 31 · 13 of 35 measured | 10.7%high reasoning | 55.4% |
| Reliably publish-ready · sales brief% of products, higher is better 28 · 7 of 35 measured | 1.8%high reasoning | 51.8% |
| DeckBench measured by Spring Prompt · 3 results | ||
| Cost per deckUS dollars, lower is better 8 · 3 of 36 measured | $0.0148 | $0.0087 |
| Deck ratingrating, higher is betterThe rating weighs each win by how strong the other deck was; the win rate counts wins against whichever decks a model happened to meet, so the two can disagree. 34 · 12 of 36 measured | 595 | 1,252 |
| Accurate decks% of tasks, higher is better 31 · 17 of 36 measured | 0.0% | 50.0% |
| ROASBench measured by Spring Prompt · 3 results | ||
| Overall scorescore out of 100, higher is better 26 · 7 of 36 measured | 37.0 | 53.4 |
| Audience scorescore out of 100, higher is better 25 · 7 of 36 measured | 55.8 | 68.3 |
| Business scorescore out of 100, higher is better 26 · 7 of 36 measured | 34.3 | 54.8 |
| SurveyBench measured by Spring Prompt · 3 results | ||
| Analysis ratingrating, higher is betterThe rating weighs each win by how strong the other analysis was; the win rate counts wins against whichever analyses a model happened to meet, so the two can disagree. 19 · 14 of 27 measured | 770 | 1,072 |
| Cost per taskUS dollars, lower is better 13 · 1 of 27 measured | $0.0924 | $0.0088 |
| Analysis scorepoints of 100, higher is better 19 · 7 of 27 measured | 63.8 | 90.8 |
| Artificial Analysis reported by Artificial Analysis · 3 results | ||
| Artificial Analysis Intelligence Indexindex score, higher is better 36 · 39 of 249 measured | 38.4 | 38.1max reasoning |
| Humanity's Last Exam% of questions, higher is better 65 · 53 of 250 measured | 35.0% | 38.5%max reasoning |
| Terminal-Bench 4.0% of tasks, higher is better 26 · 37 of 120 measured | 24.7% | 12.6%max reasoning |
Show all 89 results (71 more)
| Measure | Mistral Large 4 | GPT-6 Luna |
|---|---|---|
| BulletBench measured by Spring Prompt · 2 results | ||
| Ladder Elo · Blitz 3+2ladder Elo, higher is better 18 · 14 of 23 measured | 555 | 557 |
| Median move time · Blitz 3+2milliseconds, lower is better 3 · 19 of 23 measured | 0.8 s | 7.0 s |
| CatalogBench measured by Spring Prompt · 28 results | ||
| Channel compliance% of products, higher is better 1 · 1 of 36 measured | 100.0% | 100.0% |
| Channel rules broken% of products, lower is better 35 · 1 of 36 measured | 3.0% | 0.0% |
| Channel rules broken · sales brief% of products, lower is better 28 · 1 of 36 measured | 1.2% | 0.0% |
| Claims to check% of products, lower is better 27 · 1 of 34 measured | 0.7%high reasoning | 0.0% |
| Claims to check · sales brief% of products, lower is better 27 · 1 of 34 measured | 3.3%high reasoning | 0.0% |
| Conflicts caught% of conflicts, higher is better 1 · 1 of 36 measured | 100.0% | 100.0% |
| Content quality% of checks, higher is better 23 · 20 of 36 measured | 96.5% | 97.0% |
| Content quality · sales brief% of checks, higher is better 28 · 18 of 36 measured | 95.9% | 97.4% |
| Decision accuracy% of decisions, higher is better 33 · 16 of 36 measured | 93.1% | 97.0% |
| Failed outputs% of products, lower is better 1 · 1 of 36 measured | 0.0% | 0.0% |
| Failed outputs · sales brief% of products, lower is better 1 · 1 of 36 measured | 0.0% | 0.0% |
| Field accuracy% of missing fields, higher is better 10 · 18 of 36 measured | 94.4% | 93.7% |
| Invented values% of filled values, lower is better 32 · 18 of 36 measured | 3.5% | 1.7% |
| Missing UK information% of products, lower is better 1 · 1 of 36 measured | 0.0% | 0.0% |
| Missing UK information · sales brief% of products, lower is better 1 · 1 of 36 measured | 0.0% | 0.0% |
| Not findable% of products, lower is better 18 · 32 of 36 measured | 4.2% | 10.1% |
| Not findable · sales brief% of products, lower is better 15 · 32 of 36 measured | 3.0%high reasoning | 10.1% |
| Publish-ready listings% of products, higher is better 30 · 17 of 36 measured | 39.9%high reasoning | 67.9% |
| Publish-ready listings · sales brief% of products, higher is better 29 · 8 of 36 measured | 10.7%high reasoning | 67.9% |
| Time per productseconds, lower is better 8 · 8 of 36 measured | 8.20 | 8.20 |
| Unsupported claimsclaims per product, lower is better 26 · 13 of 36 measured | 0.91high reasoning | 0.31 |
| Unsupported claims% of products, lower is better 29 · 13 of 36 measured | 33.3%high reasoning | 8.9% |
| Unsupported claims · sales briefclaims per product, lower is better 31 · 8 of 36 measured | 3.69high reasoning | 0.35 |
| Unsupported claims · sales brief% of products, lower is better 30 · 8 of 36 measured | 80.4%high reasoning | 8.3% |
| Wrong attributes% of products, lower is better 26 · 24 of 36 measured | 19.6%high reasoning | 18.5% |
| Wrong attributes · sales brief% of products, lower is better 20 · 23 of 36 measured | 17.3%high reasoning | 19.1% |
| Wrong category or variant% of products, lower is better 5 · 7 of 36 measured | 1.2% | 1.8% |
| Wrong category or variant · sales brief% of products, lower is better 7 · 1 of 36 measured | 1.8% | 0.0% |
| DeckBench measured by Spring Prompt · 14 results | ||
| Caveat dropped% of tasks, lower is better 16 · 1 of 36 measured | 16.7% | 0.0% |
| Clean layout% of tasks, higher is better 18 · 10 of 36 measured | 33.3% | 66.7% |
| Design quality% of the maximum, higher is better 35 · 10 of 36 measured | 30.5% | 78.1% |
| Draft figure quoted% of tasks, lower is better 15 · 21 of 36 measured | 16.7% | 33.3% |
| Findings missing% of tasks, lower is better 1 · 1 of 36 measured | 0.0% | 0.0% |
| Head-to-head win rate% of comparisons, higher is betterThe rating weighs each win by how strong the other deck was; the win rate counts wins against whichever decks a model happened to meet, so the two can disagree. 35 · 9 of 36 measured | 24.7% | 70.9% |
| Layout defects% of tasks, lower is better 18 · 10 of 36 measured | 66.7% | 33.3% |
| Misleading metric used% of tasks, lower is better 1 · 1 of 36 measured | 0.0% | 0.0% |
| Presentable decks% of tasks, higher is better 17 · 17 of 36 measured | 0.0% | 0.0% |
| Recommendation late or wrong% of tasks, lower is better 33 · 17 of 36 measured | 83.3% | 16.7% |
| Slides needing work% of tasks, lower is better 36 · 5 of 36 measured | 100.0% | 16.7% |
| Time per deckseconds, lower is better 18 · 17 of 36 measured | 114 | 101 |
| Unsupported claims% of tasks, lower is better 32 · 1 of 36 measured | 50.0% | 0.0% |
| Unsupported numbers% of tasks, lower is better 34 · 1 of 36 measured | 33.3% | 0.0% |
| ROASBench measured by Spring Prompt · 6 results | ||
| Consistency scorescore out of 100, higher is better 30 · 10 of 36 measured | 16.7 | 35.8 |
| Contribution profitsimulated US dollars, higher is better 23 · 9 of 36 measured | $533,401 | $738,879 |
| Cost of a runUS dollars, lower is better 9 · 2 of 36 measured | $0.0657 | $0.0142 |
| Months over budgetmonths of 12, lower is better 1 · 1 of 36 measured | 0high reasoning | 0 |
| Planning scorescore out of 100, higher is better 7 · 20 of 36 measured | 55.0high reasoning | 54.8 |
| Return on ad spendprofit per $1 spent, higher is better 23 · 10 of 36 measured | $0.45 | $0.66 |
| SurveyBench measured by Spring Prompt · 17 results | ||
| Head-to-head win rate% of comparisons, higher is betterThe rating weighs each win by how strong the other analysis was; the win rate counts wins against whichever analyses a model happened to meet, so the two can disagree. 21 · 17 of 27 measured | 35.5% | 47.1% |
| Leading or double-barrelled questions% of tasks, lower is better 12 · 5 of 27 measured | 33.3% | 11.1% |
| Mistaken next steps% of tasks, lower is better 23 · 5 of 27 measured | 66.7% | 11.1% |
| Numbers right% of answers, higher is better 19 · 21 of 27 measured | 98.9% | 98.0% |
| Objectives unanswerable% of tasks, lower is better 18 · 1 of 27 measured | 11.1% | 0.0% |
| Precision ignored% of tasks, lower is better 23 · 1 of 27 measured | 55.6% | 0.0% |
| Research scorepoints of 100, higher is better 19 · 7 of 27 measured | 71.2 | 91.7 |
| Sound research% of tasks, higher is better 17 · 6 of 27 measured | 0.0% | 33.3% |
| Stakeholder's view presumed% of tasks, lower is better 18 · 12 of 27 measured | 22.2% | 11.1% |
| Survey plan scorepoints of 100, higher is better 19 · 9 of 27 measured | 88.4 | 94.0 |
| Survey structure% of tasks, lower is better 3 · 12 of 27 measured | 11.1% | 22.2% |
| Time per taskseconds, lower is better 25 · 12 of 27 measured | 439 | 138 |
| Traps handled% of traps, higher is better 20 · 11 of 27 measured | 78.7% | 94.2% |
| Traps missed% of tasks, lower is better 25 · 8 of 27 measured | 100.0% | 33.3% |
| Unsupported findings% of tasks, lower is better 24 · 7 of 27 measured | 100.0% | 22.2% |
| Wrong numbers% of tasks, lower is better 17 · 14 of 27 measured | 22.2% | 11.1% |
| Wrong recommendation% of tasks, lower is better 19 · 5 of 27 measured | 44.4% | 11.1% |
| Artificial Analysis reported by Artificial Analysis · 4 results | ||
| Long-context reasoning (AA-LCR)% of questions, higher is better 32 · 13 of 241 measured | 81.3% | 83.3%max reasoning |
| Output speedtokens per second, higher is better 61 · 37 of 78 measured | 60.2 | 127max reasoning |
| SciCode% of problems, higher is better 34 · 31 of 121 measured | 54.2% | 54.6%max reasoning |
| Time to first answer tokenseconds, lower is better 68 · 20 of 78 measured | 33.9 | 0.74no reasoning |
Compare them with other models
Quick answers
Is Mistral Large 4 better than GPT-6 Luna?
Pick GPT-6 Luna: it scores higher (Spring Prompt overall 38 against 24) and costs less.
Which is cheaper, Mistral Large 4 or GPT-6 Luna?
GPT-6 Luna is cheaper to run: $0.000 against $0.003 per product listing on CatalogBench (6.9 times less).
Which is faster, Mistral Large 4 or GPT-6 Luna?
GPT-6 Luna writes about 127 tokens a second on its usual API, against 60 for Mistral Large 4 (Artificial Analysis).
Which is better for product listings, Mistral Large 4 or GPT-6 Luna?
GPT-6 Luna ranks 10 of 35 models on sale for turning a sparse product feed and photos into listings that can go live, against 30 for Mistral Large 4.
Which is better for decks from an analysis, Mistral Large 4 or GPT-6 Luna?
GPT-6 Luna ranks 12 of 35 models on sale for turning a finished analysis into a deck you could present as it is, against 33 for Mistral Large 4.
Which is better for user surveys, Mistral Large 4 or GPT-6 Luna?
GPT-6 Luna ranks 14 of 26 models on sale for planning a user survey and reading its results without being misled, against 19 for Mistral Large 4.
Which is better for marketing planning, Mistral Large 4 or GPT-6 Luna?
GPT-6 Luna ranks 7 of 35 models on sale for planning a year of ad spend without overspending, against 26 for Mistral Large 4.
Which has the bigger context window, Mistral Large 4 or GPT-6 Luna?
GPT-6 Luna: 1,050,000 tokens, against 1,048,576 for Mistral Large 4.
Run Mistral Large 4 and GPT-6 Luna on your own prompt.
Benchmarks aren't your task. Test both side by side on your prompt, with every answer checked against your rules and the cost of each one.

