Compare / Gemini 4 Argon vs GPT-6 Luna
Gemini 4 Argon vs GPT-6 Luna
Which is cheaper, which is more capable, and which does better at each kind of work, from 83 results on 6 sources that measured both models.
GoogleGemini 4 Argon
OpenAIGPT-6 Luna
Gemini 4 Argon wins 3121 too close to callGPT-6 Luna wins 31
Of the 83 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.
Results as of 14 October 2026 (catalogue release 2026-10-14-7d7cf32c3f6d). See also Google vs OpenAI, every model from both labs.
Gemini 4 Argon or GPT-6 Luna? The short answer
Pick Gemini 4 Argon if quality matters most: it scores higher (Spring Prompt overall 75 against 39). Pick GPT-6 Luna if cost matters more: it costs 23 times less per product listing on CatalogBench.
- Spring Prompt overall: Gemini 4 Argon 75, GPT-6 Luna 39 (out of 100; how the score works).
- GPT-6 Luna is cheaper to run: $0.000 against $0.011 per product listing on CatalogBench (23 times less).
- Gemini 4 Argon scores higher on the Artificial Analysis Intelligence Index: 52.6 against 38.1.
- Gemini 4 Argon is newer: released 13 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 decks from an analysis and marketing planning.
- On third-party benchmarks, on the results both models have: Gemini 4 Argon does better for agents and tool use, reasoning and knowledge, writing and sticking to the facts.
Head to head
75
Spring Prompt overall
39
52.6
Intelligence Index
38.1
$4.00
Price per million tokens, blendedlower is better
$0.20
1,048,576
Context window, tokens
1,050,000
Every fact side by side
| Gemini 4 Argon | GPT-6 Luna | |
|---|---|---|
| Price per million tokens | $2.00 in · $10.00 out | $0.10 in · $0.50 out |
| Blended (3 in : 1 out) | $4.00 | $0.20 |
| Cheapest host, blended | – | $0.20 (Azure) |
| Cost per product listing (CatalogBench) | $0.011 | $0.000 |
| Cached input, per million | – | $0.010 cached input |
| Intelligence Index | 52.6 | 38.1 |
| Spring Prompt overall | 75 | 39 |
| Output speed, tokens a second | – | 128 |
| Context window | 1,048,576 tokens | 1,050,000 tokens |
| Released | 13 Oct 2026 (newer) | 22 Sep 2026 |
| Weights | Closed | Closed |
| Developer based in | the US | 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.
Product listingsTurning a sparse product feed and photos into listings that can go live
7 of 22
5 of 22
Decks from an analysisTurning a finished analysis into a deck you could present as it is
13 of 21
7 of 21
User surveysPlanning a user survey and reading its results without being misled
5 of 10
not measured
Marketing planningPlanning a year of ad spend without overspending
8 of 21
3 of 21
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls
3 of 166
46 of 166
Professional workReal tasks from banking, consulting and law, business documents and freelance projects
not measured
25 of 51
Reasoning and knowledgeHard questions across science, maths and general knowledge
3 of 200
43 of 200
WritingWhat people prefer in blind comparisons, and judged writing quality
1 of 189
61 of 189
Sticking to the factsSummarising without inventing things, and factual answers
1 of 153
95 of 153
Speed under pressureGood decisions against a real clock (fast chess)
20 of 25
19 of 25
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
- Gemini 4 Argon review: Opus-level product data at a quarter of the cost, until you ask it to sell 14 Oct 2026
- AI Checkpoint: Sol’s new fast lane 9 Oct 2026
- We Raced AI Models to Find One Small Thing in a Huge Picture. Cutting It Up More Than Doubled What They Found. 8 Oct 2026
- Claude Haiku 5.5 Is Far Better Than Haiku 4.5 at a Tenth of the Price. GPT-6 Luna Still Beats It at Business Work. 8 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 | Gemini 4 Argon | GPT-6 Luna |
|---|---|---|
| BulletBench measured by Spring Prompt · 3 results | ||
| Cost per game · Blitz 3+2US dollars, lower is better 10 · 1 of 15 measured | $0.12 | $0.0072 |
| Cost per game · Bullet 60sUS dollars, lower is better 10 · 5 of 30 measured | $0.0036 | $0.0016no reasoning |
| Ladder Elo · Bullet 60sladder Elo, higher is better 26 · 24 of 30 measured | 147 | 249no reasoning |
| CatalogBench measured by Spring Prompt · 3 results | ||
| Cost per productUS dollars, lower is better 12 · 1 of 22 measured | $0.0109 | $0.0005 |
| Reliably publish-ready% of products, higher is better 5 · 7 of 22 measured | 60.7% | 55.4% |
| Reliably publish-ready · sales brief% of products, higher is better 9 · 4 of 22 measured | 17.9% | 51.8% |
| DeckBench measured by Spring Prompt · 3 results | ||
| Cost per deckUS dollars, lower is better 10 · 1 of 21 measured | $0.11 | $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. 13 · 7 of 21 measured | 998 | 1,331 |
| Accurate decks% of tasks, higher is better 13 · 9 of 21 measured | 33.3% | 50.0% |
| ROASBench measured by Spring Prompt · 3 results | ||
| Overall scorescore out of 100, higher is better 8 · 3 of 21 measured | 47.0 | 53.4 |
| Audience scorescore out of 100, higher is better 8 · 3 of 21 measured | 65.5 | 68.3 |
| Business scorescore out of 100, higher is better 8 · 3 of 21 measured | 43.3 | 54.8 |
| Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results | ||
| Overall · TextArena rating, higher is better 1 · 78 of 387 measured | 1,525high reasoning | 1,443max reasoning |
| Overall · Text factualityArena rating, higher is better 1 · 98 of 163 measured | 1,511high reasoning | 1,430max reasoning |
| Business, management and finance · TextArena rating, higher is better 1 · 71 of 380 measured | 1,522high reasoning | 1,447max reasoning |
| Artificial Analysis reported by Artificial Analysis · 3 results | ||
| Artificial Analysis Intelligence Indexindex score, higher is better 5 · 36 of 244 measured | 52.6high reasoning | 38.1max reasoning |
| Humanity's Last Exam% of questions, higher is better 3 · 48 of 243 measured | 57.1%high reasoning | 38.5%max reasoning |
| Terminal-Bench 4.0% of tasks, higher is better 4 · 33 of 113 measured | 57.1%high reasoning | 12.6%max reasoning |
Show all 83 results (65 more)
| Measure | Gemini 4 Argon | GPT-6 Luna |
|---|---|---|
| BulletBench measured by Spring Prompt · 7 results | ||
| Games lost on time · Blitz 3+2% of games, lower is better 1 · 12 of 15 measured | 0.0% | 50.0% |
| Games lost on time · Bullet 60s% of games, lower is better 28 · 19 of 30 measured | 87.5% | 50.0%no reasoning |
| Invalid moves · Blitz 3+2% of moves, lower is better 1 · 1 of 15 measured | 0.0% | 0.0% |
| Invalid moves · Bullet 60s% of moves, lower is better 1 · 1 of 30 measured | 0.0% | 0.0%no reasoning |
| Ladder Elo · Blitz 3+2ladder Elo, higher is better 1 · 7 of 15 measured | 1,147 | 557 |
| Median move time · Blitz 3+2milliseconds, lower is better 5 · 12 of 15 measured | 1.9 s | 7.0 s |
| Median move time · Bullet 60smilliseconds, lower is better 28 · 19 of 30 measured | 2.9 s | 1.5 sno reasoning |
| CatalogBench measured by Spring Prompt · 28 results | ||
| Channel compliance% of products, higher is better 1 · 1 of 22 measured | 100.0% | 100.0% |
| Channel rules broken% of products, lower is better 1 · 1 of 22 measured | 0.0% | 0.0% |
| Channel rules broken · sales brief% of products, lower is better 1 · 1 of 22 measured | 0.0% | 0.0% |
| Claims to check% of products, lower is better 1 · 1 of 22 measured | 0.0% | 0.0% |
| Claims to check · sales brief% of products, lower is better 1 · 1 of 22 measured | 0.0% | 0.0% |
| Conflicts caught% of conflicts, higher is better 1 · 1 of 22 measured | 100.0% | 100.0% |
| Content quality% of checks, higher is better 4 · 11 of 22 measured | 98.2% | 97.0% |
| Content quality · sales brief% of checks, higher is better 4 · 10 of 22 measured | 98.4% | 97.4% |
| Decision accuracy% of decisions, higher is better 5 · 9 of 22 measured | 97.8% | 97.0% |
| Failed outputs% of products, lower is better 1 · 1 of 22 measured | 0.0% | 0.0% |
| Failed outputs · sales brief% of products, lower is better 1 · 1 of 22 measured | 0.0% | 0.0% |
| Field accuracy% of missing fields, higher is better 3 · 10 of 22 measured | 94.7% | 93.7% |
| Invented values% of filled values, lower is better 14 · 13 of 22 measured | 1.7% | 1.7% |
| Missing UK information% of products, lower is better 1 · 1 of 22 measured | 0.0% | 0.0% |
| Missing UK information · sales brief% of products, lower is better 1 · 1 of 22 measured | 0.0% | 0.0% |
| Not findable% of products, lower is better 3 · 19 of 22 measured | 2.4% | 10.1% |
| Not findable · sales brief% of products, lower is better 2 · 19 of 22 measured | 1.8% | 10.1% |
| Publish-ready listings% of products, higher is better 5 · 8 of 22 measured | 70.8% | 67.9% |
| Publish-ready listings · sales brief% of products, higher is better 8 · 4 of 22 measured | 35.1% | 67.9% |
| Time per productseconds, lower is better 3 · 5 of 22 measured | 7.30 | 8.20 |
| Unsupported claimsclaims per product, lower is better 6 · 5 of 22 measured | 0.42 | 0.31 |
| Unsupported claims% of products, lower is better 9 · 5 of 22 measured | 16.1% | 8.9% |
| Unsupported claims · sales briefclaims per product, lower is better 12 · 4 of 22 measured | 1.98 | 0.35 |
| Unsupported claims · sales brief% of products, lower is better 12 · 4 of 22 measured | 63.1% | 8.3% |
| Wrong attributes% of products, lower is better 6 · 13 of 22 measured | 15.5% | 18.5% |
| Wrong attributes · sales brief% of products, lower is better 7 · 13 of 22 measured | 15.5% | 19.1% |
| Wrong category or variant% of products, lower is better 4 · 4 of 22 measured | 1.8% | 1.8% |
| Wrong category or variant · sales brief% of products, lower is better 4 · 1 of 22 measured | 1.8% | 0.0% |
| DeckBench measured by Spring Prompt · 14 results | ||
| Caveat dropped% of tasks, lower is better 1 · 1 of 21 measured | 0.0% | 0.0% |
| Clean layout% of tasks, higher is better 13 · 6 of 21 measured | 33.3% | 66.7% |
| Design quality% of the maximum, higher is better 16 · 5 of 21 measured | 55.2% | 78.1% |
| Draft figure quoted% of tasks, lower is better 17 · 12 of 21 measured | 50.0% | 33.3% |
| Findings missing% of tasks, lower is better 1 · 1 of 21 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. 13 · 8 of 21 measured | 57.0% | 76.9% |
| Layout defects% of tasks, lower is better 13 · 6 of 21 measured | 66.7% | 33.3% |
| Misleading metric used% of tasks, lower is better 1 · 1 of 21 measured | 0.0% | 0.0% |
| Presentable decks% of tasks, higher is better 12 · 12 of 21 measured | 0.0% | 0.0% |
| Recommendation late or wrong% of tasks, lower is better 1 · 9 of 21 measured | 0.0% | 16.7% |
| Slides needing work% of tasks, lower is better 15 · 2 of 21 measured | 66.7% | 16.7% |
| Time per deckseconds, lower is better 6 · 12 of 21 measured | 67.5 | 101 |
| Unsupported claims% of tasks, lower is better 11 · 1 of 21 measured | 16.7% | 0.0% |
| Unsupported numbers% of tasks, lower is better 1 · 1 of 21 measured | 0.0% | 0.0% |
| ROASBench measured by Spring Prompt · 6 results | ||
| Consistency scorescore out of 100, higher is better 5 · 4 of 21 measured | 34.7 | 35.8 |
| Contribution profitsimulated US dollars, higher is better 5 · 4 of 21 measured | $673,561 | $738,879 |
| Cost of a runUS dollars, lower is better 11 · 1 of 21 measured | $0.24 | $0.0142 |
| Months over budgetmonths of 12, lower is better 1 · 1 of 21 measured | 0 | 0 |
| Planning scorescore out of 100, higher is better 1 · 12 of 21 measured | 55.4 | 54.8 |
| Return on ad spendprofit per $1 spent, higher is better 8 · 4 of 21 measured | $0.55 | $0.66 |
| Arena (formerly LMArena) reported by Arena (formerly LMArena) · 8 results | ||
| Confirmed task success · AgentIPS effect estimate, higher is better 3 · 29 of 48 measured | 0.15high reasoning, arena agent | -0.00max reasoning, arena agent |
| Creative writing · TextArena rating, higher is better 1 · 77 of 385 measured | 1,519high reasoning | 1,413max reasoning |
| Expert prompts · TextArena rating, higher is better 7 · 40 of 338 measured | 1,539high reasoning | 1,498max reasoning |
| Instruction following · TextArena rating, higher is better 1 · 57 of 387 measured | 1,528high reasoning | 1,448max reasoning |
| Praise over complaint · AgentIPS effect estimate, higher is better 4 · 30 of 48 measured | 0.28high reasoning, arena agent | -0.02max reasoning, arena agent |
| Steerability · AgentIPS effect estimate, higher is better 1 · 16 of 48 measured | 0.13high reasoning, arena agent | 0.04max reasoning, arena agent |
| Tool grounding · AgentIPS effect estimate, higher is better 29 · 39 of 48 measured | 0.00high reasoning, arena agent | -0.00max reasoning, arena agent |
| Writing, literature and language · TextArena rating, higher is better 1 · 79 of 386 measured | 1,522high reasoning | 1,421max reasoning |
| Artificial Analysis reported by Artificial Analysis · 2 results | ||
| Long-context reasoning (AA-LCR)% of questions, higher is better 44 · 12 of 234 measured | 79.7%high reasoning | 83.3%max reasoning |
| SciCode% of problems, higher is better 3 · 27 of 114 measured | 61.8%high reasoning | 54.6%max reasoning |
Compare them with other models
- Google vs OpenAIEvery model from both labs, best against best
- Gemini 4 Argon vs Claude Opus 5.5Most intelligent at $3–10 per million tokens
- Gemini 4 Argon vs Claude Sonnet 5.5Cheapest with similar intelligence
- Gemini 4 Argon vs GPT-6 AstraJust above overall
- Gemini 4 Argon vs Muse WonderJust below overall
- GPT-6 Luna vs Claude Haiku 5.5Most intelligent at under $0.25 per million tokens
- GPT-6 Luna vs DeepSeek-V4.1-FlashFastest with similar intelligence
- GPT-6 Luna vs Claude Opus 5.5Most intelligent available today
Quick answers
Is Gemini 4 Argon better than GPT-6 Luna?
Pick Gemini 4 Argon if quality matters most: it scores higher (Spring Prompt overall 75 against 39). Pick GPT-6 Luna if cost matters more: it costs 23 times less per product listing on CatalogBench.
Which is cheaper, Gemini 4 Argon or GPT-6 Luna?
GPT-6 Luna is cheaper to run: $0.000 against $0.011 per product listing on CatalogBench (23 times less).
Which is better for product listings, Gemini 4 Argon or GPT-6 Luna?
GPT-6 Luna ranks 5 of 22 models on sale for turning a sparse product feed and photos into listings that can go live, against 7 for Gemini 4 Argon.
Which is better for decks from an analysis, Gemini 4 Argon or GPT-6 Luna?
GPT-6 Luna ranks 7 of 21 models on sale for turning a finished analysis into a deck you could present as it is, against 13 for Gemini 4 Argon.
Which is better for marketing planning, Gemini 4 Argon or GPT-6 Luna?
GPT-6 Luna ranks 3 of 21 models on sale for planning a year of ad spend without overspending, against 8 for Gemini 4 Argon.
Which is better for agents and tool use, Gemini 4 Argon or GPT-6 Luna?
Gemini 4 Argon ranks 3 of 166 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 46 for GPT-6 Luna.
Which is better for reasoning and knowledge, Gemini 4 Argon or GPT-6 Luna?
Gemini 4 Argon ranks 3 of 200 models on sale for hard questions across science, maths and general knowledge, against 43 for GPT-6 Luna.
Which has the bigger context window, Gemini 4 Argon or GPT-6 Luna?
GPT-6 Luna: 1,050,000 tokens, against 1,048,576 for Gemini 4 Argon.
Run Gemini 4 Argon and GPT-6 Luna on your own prompt
Benchmarks aren't your data. Try both side by side in the playground, with the cost of every answer.