Compare / Claude Haiku 4.5 vs Gemini 4 Argon
Claude Haiku 4.5 vs Gemini 4 Argon
Which is cheaper, which is more capable, and which does better at each kind of work, from 99 results on 7 sources that measured both models.
AnthropicClaude Haiku 4.5
GoogleGemini 4 Argon
Claude Haiku 4.5 wins 1118 too close to callGemini 4 Argon wins 70
Of the 99 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 Anthropic vs Google, every model from both labs.
Claude Haiku 4.5 or Gemini 4 Argon? The short answer
Pick Gemini 4 Argon if quality matters most: it scores higher (Spring Prompt overall 75 against 7). Pick Claude Haiku 4.5 if cost matters more: it costs 44% less per product listing on CatalogBench.
- Spring Prompt overall: Claude Haiku 4.5 7, Gemini 4 Argon 75 (out of 100; how the score works).
- Claude Haiku 4.5 is cheaper to run: $0.006 against $0.011 per product listing on CatalogBench (44% less).
- Gemini 4 Argon scores higher on the Artificial Analysis Intelligence Index: 52.6 against 16.9.
- Gemini 4 Argon is newer: released 13 Oct 2026, against 15 Oct 2025 for Claude Haiku 4.5.
- On our own benchmarks, on the results both models have: Gemini 4 Argon does better for product listings, decks from an analysis, user surveys 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
7
Spring Prompt overall
75
16.9
Intelligence Index
52.6
$2.00
Price per million tokens, blendedlower is better
$4.00
200,000
Context window, tokens
1,048,576
Every fact side by side
| Claude Haiku 4.5 | Gemini 4 Argon | |
|---|---|---|
| Price per million tokens | $1.00 in · $5.00 out | $2.00 in · $10.00 out |
| Blended (3 in : 1 out) | $2.00 | $4.00 |
| Cheapest host, blended | $2.00 (Amazon Bedrock) | – |
| Cost per product listing (CatalogBench) | $0.006 | $0.011 |
| Cached input, per million | $0.10 cached input | – |
| Intelligence Index | 16.9 | 52.6 |
| Spring Prompt overall | 7 | 75 |
| Output speed, tokens a second | – | – |
| Context window | 200,000 tokens | 1,048,576 tokens |
| Released | 15 Oct 2025 | 13 Oct 2026 (newer) |
| 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
21 of 22
7 of 22
Decks from an analysisTurning a finished analysis into a deck you could present as it is
17 of 21
13 of 21
User surveysPlanning a user survey and reading its results without being misled
10 of 10
5 of 10
Marketing planningPlanning a year of ad spend without overspending
20 of 21
8 of 21
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls
93 of 166
3 of 166
Reasoning and knowledgeHard questions across science, maths and general knowledge
129 of 200
3 of 200
WritingWhat people prefer in blind comparisons, and judged writing quality
88 of 189
1 of 189
Sticking to the factsSummarising without inventing things, and factual answers
118 of 153
1 of 153
Speed under pressureGood decisions against a real clock (fast chess)
18 of 25
20 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
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 | Claude Haiku 4.5 | Gemini 4 Argon |
|---|---|---|
| BulletBench measured by Spring Prompt · 3 results | ||
| Cost per game · Blitz 3+2US dollars, lower is better 6 · 10 of 15 measured | $0.0318 | $0.12 |
| Cost per game · Bullet 60sUS dollars, lower is better 21 · 10 of 30 measured | $0.0163 | $0.0036 |
| Ladder Elo · Bullet 60sladder Elo, higher is better 17 · 26 of 30 measured | 445 | 147 |
| CatalogBench measured by Spring Prompt · 3 results | ||
| Cost per productUS dollars, lower is better 6 · 12 of 22 measured | $0.0061 | $0.0109 |
| Reliably publish-ready% of products, higher is better 21 · 5 of 22 measured | 1.8% | 60.7% |
| Reliably publish-ready · sales brief% of products, higher is better 20 · 9 of 22 measured | 0.0% | 17.9% |
| DeckBench measured by Spring Prompt · 3 results | ||
| Cost per deckUS dollars, lower is better 6 · 10 of 21 measured | $0.0461 | $0.11 |
| 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. 17 · 13 of 21 measured | 678 | 998 |
| Accurate decks% of tasks, higher is better 15 · 13 of 21 measured | 16.7% | 33.3% |
| ROASBench measured by Spring Prompt · 3 results | ||
| Overall scorescore out of 100, higher is better 20 · 8 of 21 measured | 17.5 | 47.0 |
| Audience scorescore out of 100, higher is better 20 · 8 of 21 measured | 30.9 | 65.5 |
| Business scorescore out of 100, higher is better 20 · 8 of 21 measured | 7.3 | 43.3 |
| 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. 10 · 5 of 10 measured | 505 | 964 |
| Cost per taskUS dollars, lower is better 3 · 6 of 10 measured | $0.0635 | $0.13 |
| Analysis scorepoints of 100, higher is better 8 · 5 of 10 measured | 56.8 | 79.6 |
| Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results | ||
| Overall · TextArena rating, higher is better 121 · 1 of 387 measured | 1,414 | 1,525high reasoning |
| Overall · Text factualityArena rating, higher is better 118 · 1 of 163 measured | 1,418 | 1,511high reasoning |
| Business, management and finance · TextArena rating, higher is better 109 · 1 of 380 measured | 1,420 | 1,522high reasoning |
| Artificial Analysis reported by Artificial Analysis · 3 results | ||
| Artificial Analysis Intelligence Indexindex score, higher is better 117 · 5 of 244 measured | 16.9thinking reasoning | 52.6high reasoning |
| Humanity's Last Exam% of questions, higher is better 150 · 3 of 243 measured | 10.4%thinking reasoning | 57.1%high reasoning |
| Terminal-Bench 4.0% of tasks, higher is better 67 · 4 of 113 measured | 0.0%thinking reasoning | 57.1%high reasoning |
Show all 99 results (78 more)
| Measure | Claude Haiku 4.5 | Gemini 4 Argon |
|---|---|---|
| BulletBench measured by Spring Prompt · 7 results | ||
| Games lost on time · Blitz 3+2% of games, lower is better 1 · 1 of 15 measured | 0.0% | 0.0% |
| Games lost on time · Bullet 60s% of games, lower is better 13 · 28 of 30 measured | 25.0% | 87.5% |
| Invalid moves · Blitz 3+2% of moves, lower is better 15 · 1 of 15 measured | 0.9% | 0.0% |
| Invalid moves · Bullet 60s% of moves, lower is better 24 · 1 of 30 measured | 0.3% | 0.0% |
| Ladder Elo · Blitz 3+2ladder Elo, higher is better 13 · 1 of 15 measured | 491 | 1,147 |
| Median move time · Blitz 3+2milliseconds, lower is better 4 · 5 of 15 measured | 1.3 s | 1.9 s |
| Median move time · Bullet 60smilliseconds, lower is better 17 · 28 of 30 measured | 1.2 s | 2.9 s |
| 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 22 · 1 of 22 measured | 4.2% | 0.0% |
| Claims to check% of products, lower is better 22 · 1 of 22 measured | 22.6% | 0.0% |
| Claims to check · sales brief% of products, lower is better 22 · 1 of 22 measured | 44.6% | 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 19 · 4 of 22 measured | 94.4% | 98.2% |
| Content quality · sales brief% of checks, higher is better 20 · 4 of 22 measured | 93.9% | 98.4% |
| Decision accuracy% of decisions, higher is better 21 · 5 of 22 measured | 90.6% | 97.8% |
| 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 10 · 3 of 22 measured | 93.7% | 94.7% |
| Invented values% of filled values, lower is better 22 · 14 of 22 measured | 6.2% | 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 18 · 3 of 22 measured | 8.9% | 2.4% |
| Not findable · sales brief% of products, lower is better 16 · 2 of 22 measured | 6.6% | 1.8% |
| Publish-ready listings% of products, higher is better 21 · 5 of 22 measured | 12.5% | 70.8% |
| Publish-ready listings · sales brief% of products, higher is better 21 · 8 of 22 measured | 3.0% | 35.1% |
| Time per productseconds, lower is better 4 · 3 of 22 measured | 8.10 | 7.30 |
| Unsupported claimsclaims per product, lower is better 22 · 6 of 22 measured | 2.69 | 0.42 |
| Unsupported claims% of products, lower is better 22 · 9 of 22 measured | 73.2% | 16.1% |
| Unsupported claims · sales briefclaims per product, lower is better 22 · 12 of 22 measured | 5.32 | 1.98 |
| Unsupported claims · sales brief% of products, lower is better 22 · 12 of 22 measured | 91.7% | 63.1% |
| Wrong attributes% of products, lower is better 22 · 6 of 22 measured | 48.2% | 15.5% |
| Wrong attributes · sales brief% of products, lower is better 22 · 7 of 22 measured | 51.2% | 15.5% |
| Wrong category or variant% of products, lower is better 20 · 4 of 22 measured | 7.7% | 1.8% |
| Wrong category or variant · sales brief% of products, lower is better 20 · 4 of 22 measured | 6.6% | 1.8% |
| 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 11 · 13 of 21 measured | 50.0% | 33.3% |
| Design quality% of the maximum, higher is better 18 · 16 of 21 measured | 47.6% | 55.2% |
| Draft figure quoted% of tasks, lower is better 17 · 17 of 21 measured | 50.0% | 50.0% |
| Findings missing% of tasks, lower is better 16 · 1 of 21 measured | 16.7% | 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. 17 · 13 of 21 measured | 29.7% | 57.0% |
| Layout defects% of tasks, lower is better 11 · 13 of 21 measured | 50.0% | 66.7% |
| 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 15 · 1 of 21 measured | 50.0% | 0.0% |
| Slides needing work% of tasks, lower is better 15 · 15 of 21 measured | 66.7% | 66.7% |
| Time per deckseconds, lower is better 4 · 6 of 21 measured | 51.7 | 67.5 |
| Unsupported claims% of tasks, lower is better 15 · 11 of 21 measured | 33.3% | 16.7% |
| 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 20 · 5 of 21 measured | 8.7 | 34.7 |
| Contribution profitsimulated US dollars, higher is better 21 · 5 of 21 measured | −$205,001 | $673,561 |
| Cost of a runUS dollars, lower is better 2 · 11 of 21 measured | $0.0298 | $0.24 |
| Months over budgetmonths of 12, lower is better 19 · 1 of 21 measured | 2 | 0 |
| Planning scorescore out of 100, higher is better 19 · 1 of 21 measured | 54.4 | 55.4 |
| Return on ad spendprofit per $1 spent, higher is better 21 · 8 of 21 measured | −$0.21 | $0.55 |
| 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. 10 · 5 of 10 measured | 18.5% | 42.2% |
| Leading or double-barrelled questions% of tasks, lower is better 8 · 8 of 10 measured | 66.7% | 66.7% |
| Mistaken next steps% of tasks, lower is better 8 · 5 of 10 measured | 55.6% | 33.3% |
| Numbers right% of answers, higher is better 10 · 7 of 10 measured | 63.7% | 97.1% |
| Objectives unanswerable% of tasks, lower is better 9 · 1 of 10 measured | 22.2% | 0.0% |
| Precision ignored% of tasks, lower is better 8 · 4 of 10 measured | 100.0% | 11.1% |
| Research scorepoints of 100, higher is better 8 · 5 of 10 measured | 64.9 | 83.3 |
| Sound research% of tasks, higher is better 5 · 5 of 10 measured | 0.0% | 0.0% |
| Stakeholder's view presumed% of tasks, lower is better 6 · 4 of 10 measured | 22.2% | 11.1% |
| Survey plan scorepoints of 100, higher is better 7 · 4 of 10 measured | 83.6 | 91.8 |
| Survey structure% of tasks, lower is better 1 · 1 of 10 measured | 0.0% | 0.0% |
| Time per taskseconds, lower is better 5 · 3 of 10 measured | 92.6 | 74.4 |
| Traps handled% of traps, higher is better 8 · 5 of 10 measured | 62.3% | 84.1% |
| Traps missed% of tasks, lower is better 6 · 5 of 10 measured | 88.9% | 77.8% |
| Unsupported findings% of tasks, lower is better 8 · 7 of 10 measured | 100.0% | 88.9% |
| Wrong numbers% of tasks, lower is better 8 · 7 of 10 measured | 100.0% | 33.3% |
| Wrong recommendation% of tasks, lower is better 7 · 3 of 10 measured | 44.4% | 11.1% |
| Arena (formerly LMArena) reported by Arena (formerly LMArena) · 4 results | ||
| Creative writing · TextArena rating, higher is better 112 · 1 of 385 measured | 1,388 | 1,519high reasoning |
| Expert prompts · TextArena rating, higher is better 97 · 7 of 338 measured | 1,451 | 1,539high reasoning |
| Instruction following · TextArena rating, higher is better 101 · 1 of 387 measured | 1,415 | 1,528high reasoning |
| Writing, literature and language · TextArena rating, higher is better 106 · 1 of 386 measured | 1,400 | 1,522high reasoning |
| Artificial Analysis reported by Artificial Analysis · 2 results | ||
| Long-context reasoning (AA-LCR)% of questions, higher is better 83 · 44 of 234 measured | 74.3%thinking reasoning | 79.7%high reasoning |
| SciCode% of problems, higher is better 74 · 3 of 114 measured | 42.2%thinking reasoning | 61.8%high reasoning |
Compare them with other models
- Anthropic vs GoogleEvery model from both labs, best against best
- Claude Haiku 4.5 vs Muse Spark 1.3Most intelligent at $1–3 per million tokens
- Claude Haiku 4.5 vs Claude Opus 5.5Most intelligent available today
- 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
Quick answers
Is Claude Haiku 4.5 better than Gemini 4 Argon?
Pick Gemini 4 Argon if quality matters most: it scores higher (Spring Prompt overall 75 against 7). Pick Claude Haiku 4.5 if cost matters more: it costs 44% less per product listing on CatalogBench.
Which is cheaper, Claude Haiku 4.5 or Gemini 4 Argon?
Claude Haiku 4.5 is cheaper to run: $0.006 against $0.011 per product listing on CatalogBench (44% less).
Which is better for product listings, Claude Haiku 4.5 or Gemini 4 Argon?
Gemini 4 Argon ranks 7 of 22 models on sale for turning a sparse product feed and photos into listings that can go live, against 21 for Claude Haiku 4.5.
Which is better for decks from an analysis, Claude Haiku 4.5 or Gemini 4 Argon?
Gemini 4 Argon ranks 13 of 21 models on sale for turning a finished analysis into a deck you could present as it is, against 17 for Claude Haiku 4.5.
Which is better for user surveys, Claude Haiku 4.5 or Gemini 4 Argon?
Gemini 4 Argon ranks 5 of 10 models on sale for planning a user survey and reading its results without being misled, against 10 for Claude Haiku 4.5.
Which is better for marketing planning, Claude Haiku 4.5 or Gemini 4 Argon?
Gemini 4 Argon ranks 8 of 21 models on sale for planning a year of ad spend without overspending, against 20 for Claude Haiku 4.5.
Which is better for agents and tool use, Claude Haiku 4.5 or Gemini 4 Argon?
Gemini 4 Argon ranks 3 of 166 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 93 for Claude Haiku 4.5.
Which has the bigger context window, Claude Haiku 4.5 or Gemini 4 Argon?
Gemini 4 Argon: 1,048,576 tokens, against 200,000 for Claude Haiku 4.5.
Run Claude Haiku 4.5 and Gemini 4 Argon on your own prompt
Benchmarks aren't your data. Try both side by side in the playground, with the cost of every answer.