Models / Mercury 3
Inception
Mercury 3
Mercury 3 is strongest at speed under pressure; worth a look for simple, high-volume work where cost matters most.
Results 20 Oct 2026 · 63 results · 4 sources
- Price per million tokens
- $0.20 in · $0.70 out
Inception's list price, 13 Oct 2026 · compare prices$0.02 cached input - Speed
- Not measured
- Intelligence Index
- Not measured
- Spring Prompt overall
- 20th of 27 score 39 of 100 · how it works
- Developer
- Inception, based in the US
Weights not published - Released
- 19 Oct 2026
1,000,000 tokens of context - Pricing and availability
- 50% off at launch ($0.10 / $0.35); no end date given
Where it stands among the models you could choose
Available models grouped by price, intelligence, speed and developer location. Retired models are excluded.
| Among | Intelligence Index | Spring Prompt overall | Price, blended | Speed |
|---|---|---|---|---|
| Models available today | – | 20th of 27Claude Opus 5.5 | $0.33 blended, 27th percentilecheapest: Mistral Nemo | – |
| Models priced $0.25–1 per million tokens | – | 4th of 5DeepSeek-V4.1-Flash | sets this group | – |
| Models from developers based in the US | – | 15th of 17Claude Opus 5.5 | $0.33 blended, 23rd percentilecheapest: Llama 3.1 8B Instruct | – |
The name under each rank is the leader of that group; green is the top quarter of the group and red the bottom quarter, by rank. Price is shown as a percentile: the share of the group that costs less, so lower is cheaper. "Sets this group" marks the measure the group is defined by. Spring Prompt overall is our score out of 100 across every source (how it works). Price is the developer's list price (or the typical OpenRouter host where we haven't read one) for three input tokens to one output; intelligence and speed are from Artificial Analysis (data sourced from Artificial Analysis); "similar intelligence" means within 4 points on its Intelligence Index. Developer locations are where each company is based, not where a model is served. Groups under 3 models are not ranked.
Measured cost per task
API cost recorded in our benchmark runs, excluding judging.
| Task | Benchmark | Cost per task |
|---|---|---|
| Deck | DeckBench | ≈ $0.012 |
| Survey plan and analysis | SurveyBench | ≈ $0.022 |
| Game of bullet chess | BulletBench | ≈ $0.002 |
How good is it, and for what?
Each use case ranks the available models its benchmarks measured, then those priced $0.25–1 per million tokens. The bar shows its percentile in that field, best to the right. Open a row for the results behind it.
Product listingsTurning a sparse product feed and photos into listings that can go live
- CatalogBench: has not measured this model.
Decks from an analysisTurning a finished analysis into a deck you could present as it is
| Benchmark | Mercury 3 | Rank among available | Best available |
|---|---|---|---|
| DeckBenchDeckBench: deck rating | 812 | 18th of 24 | GPT-6 Astra 1,577 |
User surveysPlanning a user survey and reading its results without being misled
| Benchmark | Mercury 3 | Rank among available | Best available |
|---|---|---|---|
| SurveyBenchSurveyBench: analysis rating | 752 | 9th of 13 | Claude Opus 5.5 1,499 |
Marketing planningPlanning a year of ad spend without overspending
| Benchmark | Mercury 3 | Rank among available | Best available |
|---|---|---|---|
| ROASBenchROASBench: overall score | 26.1 | 18th of 24 | GPT-6 Astra 56.6 |
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls
- tau2-bench: pending. Its latest results are dated 26 May 2026, before this model was released.
- Berkeley Function Calling Leaderboard (BFCL) V4: pending. Its latest results are dated 16 Dec 2025, before this model was released.
- OpenHands Index: pending. Its latest results are dated 30 Jun 2026, before this model was released.
- Microsoft STATE-Bench: pending. Its latest results are dated 29 May 2026, before this model was released.
- Artificial Analysis: has not measured this model.
- Vending-Bench 2: pending. Its latest results are dated 1 Oct 2026, before this model was released.
Professional workReal tasks from banking, consulting and law, business documents and freelance projects
- APEX-Agents: pending. Its latest results are dated 1 Oct 2026, before this model was released.
- GDP.pdf: pending. Its latest results are dated 1 Oct 2026, before this model was released.
- Remote Labor Index: pending. Its latest results are dated 1 Oct 2026, before this model was released.
Reasoning and knowledgeHard questions across science, maths and general knowledge
- Artificial Analysis: has not measured this model.
WritingWhat people prefer in blind comparisons, and judged writing quality
- Arena (formerly LMArena): pending. Its latest results are dated 2 Oct 2026, before this model was released.
- UGI Leaderboard: pending. Its latest results are dated 2 Oct 2026, before this model was released.
Sticking to the factsSummarising without inventing things, and factual answers
- Vectara Hallucination Leaderboard: pending. Its latest results are dated 22 Sep 2026, before this model was released.
- Arena (formerly LMArena): pending. Its latest results are dated 2 Oct 2026, before this model was released.
- SimpleQA Verified (Epoch AI): pending. Its latest results are dated 29 Sep 2026, before this model was released.
Speed under pressureGood decisions against a real clock (fast chess)
| Benchmark | Mercury 3 | Rank among available | Best available |
|---|---|---|---|
| BulletBenchBulletBench Lightning 10+1: ladder Elo | 555 | 9th of 16 | Gemini 3.5 Flash-Lite 892 |
| BulletBenchBulletBench Bullet 60s: ladder Elo | 491 | 10th of 25 | Gemini 3.8 Flash 1,068 |
Nearest alternatives
- The most capable open-weights modelMiMo-V2.6-Pro$0.54 blended · Intelligence Index 46.3
Against Mercury 3 at $0.33 blended. "Similar intelligence" means within 4 points or better.
Against the alternatives
The models you would most likely weigh it against: the leaders of the groups above.
Scroll sideways to see every alternative.
| Mercury 3 | DeepSeek-V4.1-Flash highest-scoring overall at $0.25–1 per million tokens | Claude Opus 5.5 highest-scoring overall available today | GPT-6 Luna just above overall | GLM-5.3 just below overall | |
|---|---|---|---|---|---|
| Price per million tokens | $0.33 | $0.52 | $8.00 | $0.20 | $2.15 |
| Tokens a second | – | 369 | 96 | 127 | 87 |
| Intelligence Index | – | 39.5 | 57.6 | 38.1 | 44.8 |
| Spring Prompt overall | 39 | 53 | 85 | 40 | 36 |
| Rank among available models, by use case | |||||
| Product listings | – | 10th | 9th | 5th | – |
| Decks from an analysis | 18th of 24 | – | 4th | 7th | – |
| User surveys | 9th of 13 | – | 1st | – | – |
| Marketing planning | 18th of 24 | – | 5th | 3rd | 21st |
| Agents and tool use | pending | 28th | 2nd | 51st | 13th |
| Professional work | pending | – | 3rd | 25th | 15th |
| Reasoning and knowledge | – | 40th | 1st | 45th | 29th |
| Writing | pending | 29th | 5th | 61st | 24th |
| Sticking to the facts | pending | 24th | 3rd | 95th | 54th |
| Speed under pressure | 10th of 27 | – | – | 20th | 25th |
| Side by side → | Side by side → | Side by side → | |||
Shaded figures are better than Mercury 3's. A dash means no result.
Research and examples
In the playground: our pick in 0 of 28 tests, 80% of checks passed →
- Holographic trading cards (5 of 5 checks)
- Ball bouncing in a spinning hexagon (3 of 3 checks)
- The solar system, to scale in time (3 of 3 checks)
Where it has been measured
Among models available today, mid-field on every use case measured so far. Results for agents and tool use, professional work, writing, and sticking to the facts are pending: those sources last updated before this model was released.
Not timed: Artificial Analysis, our source for speed, has not tested it
- BulletBench15 results
- DeckBench18 results
- ROASBench9 results
- SurveyBench21 results
- APEX-AgentsPending: latest results 1 Oct 2026
- Arena (formerly LMArena)Pending: latest results 2 Oct 2026
- Berkeley Function Calling Leaderboard (BFCL) V4Pending: latest results 16 Dec 2025
- GDP.pdfPending: latest results 1 Oct 2026
- Microsoft STATE-BenchPending: latest results 1 Jul 2026
- OpenHands IndexPending: latest results 30 Jun 2026
- Remote Labor IndexPending: latest results 1 Oct 2026
- SimpleQA Verified (Epoch AI)Pending: latest results 29 Sep 2026
- tau2-benchPending: latest results 3 Aug 2026
- UGI LeaderboardPending: latest results 2 Oct 2026
- Vectara Hallucination LeaderboardPending: latest results 22 Sep 2026
- Vending-Bench 2Pending: latest results 1 Oct 2026
- CatalogBenchNot measured
- Artificial AnalysisNot measured
Every result
63 published results from 4 sources, each in the source's own units, with the configuration that produced it.
BulletBench · measured by Spring Prompt · 15 results
| Measure | Value | Rank | Configuration | Dated |
|---|---|---|---|---|
| BulletBench Bullet 60s: cost per gameUS dollars, lower is better | $0.0032 | 11 of 40 | Mercury 3 (low reasoning) low reasoning |
19 Oct 2026 |
| BulletBench Bullet 60s: cost per gameUS dollars, lower is better | $0.0020 | 7 of 40 | Mercury 3 (no reasoning) none reasoning |
19 Oct 2026 |
| BulletBench Bullet 60s: games lost on time% of games, lower is better | 87.5% | 34 of 40 | Mercury 3 (low reasoning) low reasoning |
19 Oct 2026 |
| BulletBench Bullet 60s: games lost on time% of games, lower is better | 0.0% | 1 of 40 | Mercury 3 (no reasoning) none reasoning |
19 Oct 2026 |
| BulletBench Bullet 60s: invalid moves% of moves, lower is better | 0.0% | 1 of 40 | Mercury 3 (low reasoning) low reasoning |
19 Oct 2026 |
| BulletBench Bullet 60s: invalid moves% of moves, lower is better | 0.0% | 1 of 40 | Mercury 3 (no reasoning) none reasoning |
19 Oct 2026 |
| BulletBench Bullet 60s: ladder Eloladder Elo, higher is better | 147 0–381 |
32 of 40 | Mercury 3 (low reasoning) low reasoning |
19 Oct 2026 |
| BulletBench Bullet 60s: ladder Eloladder Elo, higher is better | 491 299–659 |
17 of 40 | Mercury 3 (no reasoning) none reasoning |
19 Oct 2026 |
| BulletBench Bullet 60s: median move timemilliseconds, lower is better | 4.2 s | 39 of 40 | Mercury 3 (low reasoning) low reasoning |
19 Oct 2026 |
| BulletBench Bullet 60s: median move timemilliseconds, lower is better | 0.4 s | 3 of 40 | Mercury 3 (no reasoning) none reasoning |
19 Oct 2026 |
| BulletBench Lightning 10+1: cost per gameUS dollars, lower is better | $0.0010 | 5 of 26 | Mercury 3 (no reasoning) none reasoning |
19 Oct 2026 |
| BulletBench Lightning 10+1: games lost on time% of games, lower is better | 0.0% | 1 of 26 | Mercury 3 (no reasoning) none reasoning |
19 Oct 2026 |
| BulletBench Lightning 10+1: invalid moves% of moves, lower is better | 0.4% | 20 of 26 | Mercury 3 (no reasoning) none reasoning |
19 Oct 2026 |
| BulletBench Lightning 10+1: ladder Eloladder Elo, higher is better | 555 263–727 |
12 of 26 | Mercury 3 (no reasoning) none reasoning |
19 Oct 2026 |
| BulletBench Lightning 10+1: median move timemilliseconds, lower is better | 0.4 s | 3 of 26 | Mercury 3 (no reasoning) none reasoning |
19 Oct 2026 |
Not shown: Not your cost: prices are those charged on the run date.
DeckBench · measured by Spring Prompt · 18 results
| Measure | Value | Rank | Configuration | Dated |
|---|---|---|---|---|
| DeckBench: accurate decks% of tasks, higher is better | 0.0% | 20 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: caveat dropped% of tasks, lower is better | 16.7% | 12 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: clean layout% of tasks, higher is better | 66.7% | 6 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: cost per deckUS dollars, lower is better | $0.0116 | 4 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: deck ratingrating, higher is better | 812 707–913 |
18 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: design quality% of the maximum, higher is better | 58.5% | 16 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: draft figure quoted% of tasks, lower is better | 33.3% | 12 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: findings missing% of tasks, lower is better | 16.7% | 17 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: head-to-head win rate% of comparisons, higher is better | 44.0% | 16 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: layout defects% of tasks, lower is better | 33.3% | 6 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: misleading metric used% of tasks, lower is better | 16.7% | 22 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: output tokens per decktokens, lower is better | 32,448 | 2 of 3 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: presentable decks% of tasks, higher is better | 0.0% | 12 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: recommendation late or wrong% of tasks, lower is better | 83.3% | 22 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: slides needing work% of tasks, lower is better | 33.3% | 7 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: time per deckseconds, lower is better | 83.3 | 9 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: unsupported claims% of tasks, lower is better | 16.7% | 11 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| DeckBench: unsupported numbers% of tasks, lower is better | 16.7% | 20 of 24 | Mercury 3 provider default reasoning |
19 Oct 2026 |
Not shown: Not your analysis or your brand: invented companies, six public tasks, one deck per model per task at the provider's default reasoning setting.
ROASBench · measured by Spring Prompt · 9 results
| Measure | Value | Rank | Configuration | Dated |
|---|---|---|---|---|
| ROASBench: audience scorescore out of 100, higher is better | 43.2 | 22 of 25 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| ROASBench: business scorescore out of 100, higher is better | 17.6 | 19 of 25 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| ROASBench: consistency scorescore out of 100, higher is better | 15.1 | 19 of 25 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| ROASBench: contribution profitsimulated US dollars, higher is better | $104,868 | 19 of 25 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| ROASBench: cost of a runUS dollars, lower is better | $0.0533 | 6 of 25 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| ROASBench: months over budgetmonths of 12, lower is better | 0 | 1 of 25 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| ROASBench: overall scorescore out of 100, higher is better | 26.1 | 19 of 25 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| ROASBench: planning scorescore out of 100, higher is better | 53.8 | 24 of 25 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| ROASBench: return on ad spendprofit per $1 spent, higher is better | $0.10 | 19 of 25 | Mercury 3 provider default reasoning |
19 Oct 2026 |
Not shown: Not real ad performance: the market is a deterministic simulation of one invented skincare brand.
SurveyBench · measured by Spring Prompt · 21 results
| Measure | Value | Rank | Configuration | Dated |
|---|---|---|---|---|
| SurveyBench: analysis ratingrating, higher is better | 752 643–847 |
9 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: analysis scorepoints of 100, higher is better | 42.2 | 12 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: cost per taskUS dollars, lower is better | $0.0216 | 3 of 12 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: head-to-head win rate% of comparisons, higher is better | 48.6% | 6 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: leading or double-barrelled questions% of tasks, lower is better | 55.6% | 8 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: mistaken next steps% of tasks, lower is better | 22.2% | 5 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: numbers right% of answers, higher is better | 63.3% | 13 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: objectives unanswerable% of tasks, lower is better | 0.0% | 1 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: output tokens per tasktokens, lower is better | 57,343 | 2 of 2 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: precision ignored% of tasks, lower is better | 0.0% | 1 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: research scorepoints of 100, higher is better | 56.8 | 11 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: sound research% of tasks, higher is better | 0.0% | 6 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: stakeholder's view presumed% of tasks, lower is better | 22.2% | 7 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: survey plan scorepoints of 100, higher is better | 90.8 | 6 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: survey structure% of tasks, lower is better | 22.2% | 7 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: time per taskseconds, lower is better | 194 | 10 of 12 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: traps handled% of traps, higher is better | 60.7% | 11 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: traps missed% of tasks, lower is better | 88.9% | 7 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: unsupported findings% of tasks, lower is better | 44.4% | 4 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: wrong numbers% of tasks, lower is better | 44.4% | 9 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
| SurveyBench: wrong recommendation% of tasks, lower is better | 77.8% | 12 of 13 | Mercury 3 provider default reasoning |
19 Oct 2026 |
Not shown: Not your customers or your data: invented organisations with simulated respondents, nine public tasks, one run per model per task at the provider's default reasoning setting.