InceptionMercury 2.5
Mercury 2.5 is an Inception proprietary reasoning model, released 8 Sept 2026. Its best configuration ranks #520 of 785 on the BenchLeader Index at 45.3 ±7.2, in the lower half. The ± is the point: 409 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in long context (61) and lowest in maths (31). At $0.375 per million tokens blended it is cheaper than most ranked models. Output speed of 742 tokens per second puts it in the fastest quarter, with a first answer in 3.8 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 8 Sept 2026.
- Blended price
- $0.375/M
- $0.250 in · $0.750 out
- Output speed
- 742 tok/s
- measured by Artificial Analysis
- First answer
- 3.75 s
- first token 1.00 s
- Context
- 260k
- Full answer
- 4.43 s
- median, reasoning included
- Cost per run
- $0.117
- one full Intelligence Index run
- Released
- 8 Sept 2026
- Overall index45
- Reasoning41
- Coding38
- Maths31
- Knowledge44
- Long context61
- Composite43
Versions
Inception has shipped 3 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| NewMercury 3high | 19 Oct 2026 | – | – |
| Mercury 2.5this page | 8 Sept 2026 | 45.3 | #520 |
| Mercury 2 | 20 Feb 2026 | 46.9 | #468 |
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| high | – | – | 742 tok/s | 3.75 s | $0.0003 | Agents & tools 25 · Coding 16 · Knowledge 57 |
| not statedbest | 45.3 | #520 | 742 tok/s | 3.75 s | $0.0003 | Coding 38 · Composite 43 · Knowledge 44 · Long context 61 · Maths 31 · Reasoning 41 |
How its index has moved
23 Sept 2026 to 23 Oct 2026The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Humanity's Last Exam (AA)not in index | 11.8%#290 | Artificial Analysis | |
| CritPt | 0.0%#315 | Artificial Analysis |
Coding
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| SciCodeeffort not stated | – | 38.5%#175 | SciCode | |
| SciCode (AA)not in indexeffort not stated | – | 38.5%#186 | Artificial Analysis | |
| IOI | 2.4%#43 | – | Vals AI | |
| Code Migrationnot in index | 4.5%#71 | – | Vals AI | |
| Vibe Code Bench v1.1not in index | 3.7%#103 | – | Vals AI | |
| ALE-Benchnot in index | 301.6#118 | – | Epoch AI Benchmarking Hub | |
| Terminal-Bench 4.0 (AA)not in indexeffort not stated | – | 0.0%#152 | Artificial Analysis |
Agents & tools
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| Terminal-Bench 2.1 (Vals) | 34.5%#68 | – | Vals AI | |
| GDPval-AA v2.1not in indexeffort not stated | – | 0.0%#124 | Artificial Analysis | |
| CyberBenchnot in index | 45.5%#38 | – | Vals AI | |
| Finance Agent v2not in index | 18.9%#74 | – | Vals AI | |
| Harvey's Legal Agent Benchmark v1not in index | 0.0%#61 | – | Vals AI | |
| Legal Research Benchnot in index | 4.3%#75 | – | Vals AI | |
| SkillsBenchnot in index | 18.1%#33 | – | Vals AI | |
| Tax Agent Benchnot in index | 2.5%#71 | – | Vals AI | |
| Terminal-Bench 4.0 (Vals)not in index | 0.0%#50 | – | Vals AI | |
| Terminal-Bench Sciencenot in index | 0.0%#38 | – | Vals AI | |
| AutomationBenchnot in indexeffort not stated | – | 2.4%#179 | Artificial Analysis | |
| GDP.pdf (AA)not in indexeffort not stated | – | 3.6%#177 | Artificial Analysis |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| ProofBench | 3.0%#74 | Vals AI |
Knowledge
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| AA-Omniscienceeffort not stated | – | -39.5#319 | Artificial Analysis | |
| LegalBench | 83.1%#60 | – | Vals AI | |
| Excel Modeling Benchmarknot in index | 8.8%#72 | – | Vals AI | |
| MedCodenot in index | 31.3%#91 | – | Vals AI | |
| MedScribenot in index | 55.1%#102 | – | Vals AI | |
| Public Benefits Bench v1.1not in index | 45.3%#45 | – | Vals AI | |
| AA-Omniscience: accuracynot in indexeffort not stated | – | 22.6%#271 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in indexeffort not stated | – | 19.7%#284 | Artificial Analysis |
Long context
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-LCR | 71.7%#196 | Artificial Analysis |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA Intelligence Index v4.3.2 | 12.3#335 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| Inception | 236 tok/s | 1.00 s | $0.040 | $0.150 | 260k | – |
Price history
The price above, per 1M tokens, every day it changed.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.020 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0003 | $0.0003 | 4.2 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0035 | $0.0014 | 4.6 s |
| Code edit | 6,000 / 1,500 | $0.0026 | $0.0016 | 5.8 s |
| Agentic coding session | 60,000 / 4,000 | $0.018 | $0.0077 | 9.1 s |
| Structured extraction | 2,000 / 200 | $0.0007 | $0.0003 | 4.0 s |
More from Inception
See also
Questions people ask
- What are Mercury 2.5's benchmark scores?
- As of 24 Oct 2026, Mercury 2.5 scores 45.3 ±7.2 on the BenchLeader Index, strongest in long context (61) and weakest in maths (31). That rests on 6 independent benchmark results, and every score on this page links to its publisher.
- How does Mercury 2.5 rank?
- Mercury 2.5 is #520 of 785 ranked configurations on 24 Oct 2026. 409 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Mercury 2.5 cost per 1M tokens?
- As of 24 Oct 2026, Mercury 2.5 costs $0.250 per 1M input tokens and $0.750 per 1M output tokens, $0.375 blended (three input tokens per output token). The list price last changed on 6 Oct 2026.
- When was Mercury 2.5 released?
- Inception released Mercury 2.5 on 8 Sept 2026. BenchLeader has results for it at 2 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Mercury 2.5: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/mercury-2-5.