Compare / Gemini 3.8 Flash vs Gemini 4 Carbon
Gemini 3.8 Flash vs Gemini 4 Carbon
Which is cheaper, which is more capable, and which does better at each kind of work, from 88 results on 5 sources that measured both models.
GoogleGemini 3.8 Flash
GoogleGemini 4 Carbon
Gemini 3.8 Flash wins 2616 too close to callGemini 4 Carbon wins 46
Of the 88 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.
Results as of 11 November 2026 (catalogue release 2026-11-11-7ba7a6de390d).
Gemini 3.8 Flash or Gemini 4 Carbon? The short answer
They score about the same, so pick Gemini 3.8 Flash: it costs 2.1 times less per product listing on CatalogBench.
- Spring Prompt overall: Gemini 3.8 Flash 58, Gemini 4 Carbon 54 (out of 100; how the score works).
- Gemini 3.8 Flash is cheaper to run: $0.010 against $0.021 per product listing on CatalogBench (2.1 times less).
- Gemini 4 Carbon is newer: released 10 Nov 2026, against 2 Sep 2026 for Gemini 3.8 Flash.
- On our own benchmarks, on the results both models have: Gemini 3.8 Flash does better for marketing planning and speed under pressure; Gemini 4 Carbon for product listings, decks from an analysis and user surveys.
Head to head
58
Spring Prompt overall
54
$3.00
Price per million tokens, blendedlower is better
$8.00
1,048,576
Context window, tokens
1,048,576
Every fact side by side
| Gemini 3.8 Flash | Gemini 4 Carbon | |
|---|---|---|
| Price per million tokens | $1.50 in · $7.50 out | $4.00 in · $20.00 out |
| Blended (3 in : 1 out) | $3.00 | $8.00 |
| Cheapest host, blended | $1.50 (Google) | – |
| Cost per product listing (CatalogBench) | $0.010 | $0.021 |
| Cached input, per million | – | $0.20 cached input |
| Intelligence Index | 40.9 | – |
| Spring Prompt overall | 58 | 54 |
| Output speed, tokens a second | 141 | – |
| Context window | 1,048,576 tokens | 1,048,576 tokens |
| Released | 2 Sep 2026 | 10 Nov 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
19 of 30
13 of 30
Decks from an analysisTurning a finished analysis into a deck you could present as it is
24 of 30
20 of 30
User surveysPlanning a user survey and reading its results without being misled
14 of 21
11 of 21
Marketing planningPlanning a year of ad spend without overspending
11 of 30
12 of 30
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls
35 of 159
not measured
Professional workReal tasks from banking, consulting and law, business documents and freelance projects
11 of 51
not measured
Reasoning and knowledgeHard questions across science, maths and general knowledge
15 of 187
not measured
WritingWhat people prefer in blind comparisons, and judged writing quality
3 of 169
not measured
Sticking to the factsSummarising without inventing things, and factual answers
12 of 139
not measured
Speed under pressureGood decisions against a real clock (fast chess)
11 of 33
14 of 33
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
- Muse Ember review: GPT-6 Luna's price, better at analysis, and no Gemini Flash against a clock 10 Nov 2026
- Kimi K3.1 Pro first look: the best open-weight model on our board, and very slow 5 Nov 2026
- DeepSeek-V4.1-Flash-1017 review: better at analysis, much worse at sticking to the product data 19 Oct 2026
- AI Checkpoint: Argon matches Opus, until it sells 16 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 3.8 Flash | Gemini 4 Carbon |
|---|---|---|
| BulletBench measured by Spring Prompt · 3 results | ||
| Cost per game · Blitz 3+2US dollars, lower is better 10 · 7 of 18 measured | $0.0795 | $0.0534 |
| Cost per game · Bullet 60sUS dollars, lower is better 24 · 27 of 39 measured | $0.0154 | $0.0167low reasoning |
| Ladder Elo · Bullet 60sladder Elo, higher is better 1 · 24 of 39 measured | 1,068low reasoning | 358low reasoning |
| CatalogBench measured by Spring Prompt · 3 results | ||
| Cost per productUS dollars, lower is better 17 · 21 of 31 measured | $0.0103 | $0.0213 |
| Reliably publish-ready% of products, higher is better 18 · 11 of 30 measured | 35.7% | 53.6% |
| Reliably publish-ready · sales brief% of products, higher is better 21 · 15 of 30 measured | 3.6% | 14.3% |
| DeckBench measured by Spring Prompt · 3 results | ||
| Cost per deckUS dollars, lower is better 15 · 22 of 31 measured | $0.0613 | $0.16 |
| 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. 24 · 20 of 31 measured | 807 | 1,021 |
| Accurate decks% of tasks, higher is better 22 · 18 of 31 measured | 16.7% | 33.3% |
| ROASBench measured by Spring Prompt · 3 results | ||
| Overall scorescore out of 100, higher is better 11 · 12 of 31 measured | 51.4 | 51.2 |
| Audience scorescore out of 100, higher is better 10 · 11 of 31 measured | 67.1 | 66.4 |
| Business scorescore out of 100, higher is better 11 · 12 of 31 measured | 51.0 | 50.1 |
| 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. 14 · 11 of 22 measured | 898 | 1,076 |
| Cost per taskUS dollars, lower is better 14 · 17 of 22 measured | $0.0943 | $0.31 |
| Analysis scorepoints of 100, higher is better 13 · 10 of 22 measured | 74.7 | 84.9 |
Show all 88 results (73 more)
| Measure | Gemini 3.8 Flash | Gemini 4 Carbon |
|---|---|---|
| BulletBench measured by Spring Prompt · 7 results | ||
| Games lost on time · Blitz 3+2% of games, lower is better 7 · 16 of 18 measured | 12.5% | 62.5% |
| Games lost on time · Bullet 60s% of games, lower is better 21 · 30 of 39 measured | 50.0%low reasoning | 75.0%low reasoning |
| Invalid moves · Blitz 3+2% of moves, lower is better 1 · 1 of 18 measured | 0.0% | 0.0% |
| Invalid moves · Bullet 60s% of moves, lower is better 1 · 1 of 39 measured | 0.0% | 0.0%low reasoning |
| Ladder Elo · Blitz 3+2ladder Elo, higher is better 2 · 5 of 18 measured | 1,126 | 848low reasoning |
| Median move time · Blitz 3+2milliseconds, lower is better 9 · 8 of 18 measured | 4.0 s | 4.0 slow reasoning |
| Median move time · Bullet 60smilliseconds, lower is better 27 · 36 of 39 measured | 1.8 slow reasoning | 3.2 slow reasoning |
| CatalogBench measured by Spring Prompt · 28 results | ||
| Channel compliance% of products, higher is better 23 · 1 of 31 measured | 99.4% | 100.0% |
| Channel rules broken% of products, lower is better 18 · 1 of 31 measured | 0.6% | 0.0% |
| Channel rules broken · sales brief% of products, lower is better 17 · 1 of 31 measured | 0.6% | 0.0% |
| Claims to check% of products, lower is better 1 · 1 of 29 measured | 0.0% | 0.0% |
| Claims to check · sales brief% of products, lower is better 24 · 1 of 29 measured | 5.4% | 0.0% |
| Conflicts caught% of conflicts, higher is better 1 · 1 of 31 measured | 100.0% | 100.0% |
| Content quality% of checks, higher is better 9 · 14 of 31 measured | 97.7% | 97.3% |
| Content quality · sales brief% of checks, higher is better 13 · 6 of 31 measured | 97.6% | 98.4% |
| Decision accuracy% of decisions, higher is better 17 · 12 of 31 measured | 96.4% | 97.0% |
| Failed outputs% of products, lower is better 1 · 27 of 31 measured | 0.0% | 1.2% |
| Failed outputs · sales brief% of products, lower is better 1 · 1 of 31 measured | 0.0% | 0.0% |
| Field accuracy% of missing fields, higher is better 23 · 5 of 31 measured | 92.7% | 94.9% |
| Invented values% of filled values, lower is better 3 · 1 of 31 measured | 0.3% | 0.0% |
| Missing UK information% of products, lower is better 1 · 1 of 31 measured | 0.0% | 0.0% |
| Missing UK information · sales brief% of products, lower is better 1 · 1 of 31 measured | 0.0% | 0.0% |
| Not findable% of products, lower is better 25 · 5 of 31 measured | 8.3% | 1.8% |
| Not findable · sales brief% of products, lower is better 26 · 11 of 31 measured | 8.9% | 2.4% |
| Publish-ready listings% of products, higher is better 19 · 12 of 31 measured | 54.8% | 69.6% |
| Publish-ready listings · sales brief% of products, higher is better 21 · 16 of 31 measured | 16.7% | 31.0% |
| Time per productseconds, lower is better 19 · 26 of 31 measured | 18.7 | 45.3 |
| Unsupported claims% of products, lower is better 22 · 16 of 31 measured | 31.6% | 17.3% |
| Unsupported claimsclaims per product, lower is better 20 · 10 of 31 measured | 0.79 | 0.36 |
| Unsupported claims · sales briefclaims per product, lower is better 22 · 18 of 31 measured | 3.08 | 2.10 |
| Unsupported claims · sales brief% of products, lower is better 26 · 19 of 31 measured | 81.6% | 65.5% |
| Wrong attributes% of products, lower is better 10 · 2 of 31 measured | 15.5% | 11.3% |
| Wrong attributes · sales brief% of products, lower is better 2 · 5 of 31 measured | 11.9% | 14.3% |
| Wrong category or variant% of products, lower is better 25 · 2 of 31 measured | 3.6% | 0.6% |
| Wrong category or variant · sales brief% of products, lower is better 21 · 6 of 31 measured | 3.0% | 1.8% |
| DeckBench measured by Spring Prompt · 14 results | ||
| Caveat dropped% of tasks, lower is better 1 · 1 of 31 measured | 0.0% | 0.0% |
| Clean layout% of tasks, higher is better 24 · 16 of 31 measured | 16.7% | 33.3% |
| Design quality% of the maximum, higher is better 17 · 20 of 31 measured | 61.7% | 59.8% |
| Draft figure quoted% of tasks, lower is better 24 · 17 of 31 measured | 50.0% | 33.3% |
| Findings missing% of tasks, lower is better 1 · 1 of 31 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. 23 · 24 of 31 measured | 39.6% | 39.4% |
| Layout defects% of tasks, lower is better 24 · 16 of 31 measured | 83.3% | 66.7% |
| Misleading metric used% of tasks, lower is better 1 · 1 of 31 measured | 0.0% | 0.0% |
| Presentable decks% of tasks, higher is better 15 · 12 of 31 measured | 0.0% | 16.7% |
| Recommendation late or wrong% of tasks, lower is better 18 · 1 of 31 measured | 33.3% | 0.0% |
| Slides needing work% of tasks, lower is better 9 · 22 of 31 measured | 33.3% | 66.7% |
| Time per deckseconds, lower is better 13 · 24 of 31 measured | 93.0 | 228 |
| Unsupported claims% of tasks, lower is better 28 · 23 of 31 measured | 50.0% | 33.3% |
| Unsupported numbers% of tasks, lower is better 25 · 1 of 31 measured | 16.7% | 0.0% |
| ROASBench measured by Spring Prompt · 6 results | ||
| Consistency scorescore out of 100, higher is better 6 · 5 of 31 measured | 36.4 | 38.1 |
| Contribution profitsimulated US dollars, higher is better 5 · 6 of 31 measured | $842,432 | $789,366 |
| Cost of a runUS dollars, lower is better 15 · 23 of 31 measured | $0.16 | $0.52 |
| Months over budgetmonths of 12, lower is better 1 · 1 of 31 measured | 0 | 0 |
| Planning scorescore out of 100, higher is better 20 · 6 of 31 measured | 54.7 | 55.0 |
| Return on ad spendprofit per $1 spent, higher is better 11 · 9 of 31 measured | $0.65 | $0.65 |
| SurveyBench measured by Spring Prompt · 18 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. 5 · 12 of 22 measured | 65.0% | 49.4% |
| Leading or double-barrelled questions% of tasks, lower is better 19 · 16 of 22 measured | 66.7% | 55.6% |
| Mistaken next steps% of tasks, lower is better 5 · 9 of 22 measured | 11.1% | 22.2% |
| Numbers right% of answers, higher is better 1 · 1 of 22 measured | 100.0% | 100.0% |
| Objectives unanswerable% of tasks, lower is better 1 · 1 of 22 measured | 0.0% | 0.0% |
| Output tokens per tasktokens, lower is better 3 · 5 of 12 measured | 21,981 | 27,760 |
| Precision ignored% of tasks, lower is better 16 · 11 of 22 measured | 44.4% | 11.1% |
| Research scorepoints of 100, higher is better 13 · 10 of 22 measured | 77.6 | 87.0 |
| Sound research% of tasks, higher is better 13 · 10 of 22 measured | 0.0% | 11.1% |
| Stakeholder's view presumed% of tasks, lower is better 14 · 9 of 22 measured | 22.2% | 11.1% |
| Survey plan scorepoints of 100, higher is better 16 · 11 of 22 measured | 84.4 | 92.0 |
| Survey structure% of tasks, lower is better 10 · 3 of 22 measured | 22.2% | 11.1% |
| Time per taskseconds, lower is better 8 · 19 of 22 measured | 131 | 390 |
| Traps handled% of traps, higher is better 16 · 11 of 22 measured | 78.3% | 92.8% |
| Traps missed% of tasks, lower is better 14 · 10 of 22 measured | 88.9% | 44.4% |
| Unsupported findings% of tasks, lower is better 12 · 12 of 22 measured | 77.8% | 77.8% |
| Wrong numbers% of tasks, lower is better 1 · 1 of 22 measured | 0.0% | 0.0% |
| Wrong recommendation% of tasks, lower is better 1 · 12 of 22 measured | 0.0% | 22.2% |
Compare them with other models
- Gemini 3.8 Flash vs Claude Opus 5.5Most intelligent at $3–10 per million tokens
- Gemini 3.8 Flash vs DeepSeek-V4.1-FlashFastest with similar intelligence
- Gemini 3.8 Flash vs Claude Fable 5.2Highest-scoring overall available today
- Gemini 4 Carbon vs Claude Opus 5.5Highest-scoring overall at $3–10 per million tokens
- Gemini 4 Carbon vs Claude Fable 5.2Highest-scoring overall available today
Quick answers
Is Gemini 3.8 Flash better than Gemini 4 Carbon?
They score about the same, so pick Gemini 3.8 Flash: it costs 2.1 times less per product listing on CatalogBench.
Which is cheaper, Gemini 3.8 Flash or Gemini 4 Carbon?
Gemini 3.8 Flash is cheaper to run: $0.010 against $0.021 per product listing on CatalogBench (2.1 times less).
Which is better for product listings, Gemini 3.8 Flash or Gemini 4 Carbon?
Gemini 4 Carbon ranks 13 of 30 models on sale for turning a sparse product feed and photos into listings that can go live, against 19 for Gemini 3.8 Flash.
Which is better for decks from an analysis, Gemini 3.8 Flash or Gemini 4 Carbon?
Gemini 4 Carbon ranks 20 of 30 models on sale for turning a finished analysis into a deck you could present as it is, against 24 for Gemini 3.8 Flash.
Which is better for user surveys, Gemini 3.8 Flash or Gemini 4 Carbon?
Gemini 4 Carbon ranks 11 of 21 models on sale for planning a user survey and reading its results without being misled, against 14 for Gemini 3.8 Flash.
Which is better for marketing planning, Gemini 3.8 Flash or Gemini 4 Carbon?
Gemini 3.8 Flash ranks 11 of 30 models on sale for planning a year of ad spend without overspending, against 12 for Gemini 4 Carbon.
Which is better for speed under pressure, Gemini 3.8 Flash or Gemini 4 Carbon?
Gemini 3.8 Flash ranks 11 of 33 models on sale for good decisions against a real clock (fast chess), against 14 for Gemini 4 Carbon.
Run Gemini 3.8 Flash and Gemini 4 Carbon 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.