Across 4 comparable benchmarks, GPT-6.2 Astra has better reported results on 4 and worse on 0. 1 further shared result is listed below, each with the reason it is not counted.
Moderate evidence4 of 5 shared results countedHow we compare
Price basis: Listed rates vs standard rates. The price section below shows the applicable periods.
| Fact | GPT-6.2 Astra | Mercury 3 |
|---|---|---|
| Context window | 1.05M tokens | 1M tokens |
| Input / 1M | $8.00 | $0.100 |
| Output / 1M | $40.00 | $0.350 |
| Blended (4:1) | $8.93 | $0.085 |
| Status | Available | Available |
| Released | November 17, 2026 | October 19, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 2 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
GPT-6.2 Astra ahead on 1 of 1 counted
GPT-6.2 Astra ahead on 2 of 2 counted
No shared comparable results for Abstract reasoning, Mathematics, Science & health, Computer & tool use, Research & search, Long context, Multimodal, Cybersecurity.
Published scores, with matching modes where available. Gaps show GPT-6.2 Astra relative to Mercury 3. Reading the results
| Benchmark | GPT-6.2 Astra | Mercury 3 | Gap |
|---|---|---|---|
| Terminal-Bench 4.0Agentic Development · Resolution rate (%)Not counted · Different or unrecorded test environments — excluded from the tally | 64.1Extra-High · tools· legacy-default | 3.0≠Unspecified Reasoning · tools· vals-mini-swe | —#5 vs #70 of 90 |
| Vibe Code Bench v1.1Code Generation & Editing · Pass rate (%) | 89.9Max · tools | 60.9≠Thinking High · tools | +29.0 pts#9 vs #59 of 121 |
| Legal Research BenchLegal · Score | 42.8Max · tools + web | 28.9≠Thinking High · tools | 1.48×#29 vs #58 of 91 |
| Tax Agent BenchFinance · Accuracy (%, rubric × citation quality) | 21.8Max · tools + web | 18.2≠Thinking High · tools | +3.7 pts#52 vs #56 of 74 |
| Vals Index v2.2Capability Indices · GDP-weighted accuracy (%) | 67.7Max · tools | 43.7≠Thinking High · tools | +24.0 pts#7 vs #32 of 36 |
Agentic Development · Resolution rate (%)
Not counted · Different or unrecorded test environments — excluded from the tally
Code Generation & Editing · Pass rate (%)
Legal · Score
Finance · Accuracy (%, rubric × citation quality)
Capability Indices · GDP-weighted accuracy (%)
A dash means no result is available.
| Benchmark | GPT-6.2 Astra | Mercury 3 |
|---|---|---|
| FrontierMath Erdős | 10.3 | — |
| FrontierMath Tier 4 v2 | 100.0 | — |
| FrontierMath v2 | 92.6 | — |
| GPQA Diamond | 97.0 | — |
| HLE-Diamond | 72.2 | — |
| MysteryMechanism | 53.2 | — |
| Terminal-Bench-Science 0.1 | 78.6 | — |
| FrontierCode 2.0 | 49.0 | — |
| AutomationBench-AA | 77.2 | — |
| AA Intelligence Index v5.0.1 | 62.1 | — |
| Agent Arena Index | 1291 | — |
| ECI | 174 | — |
| Vals RSI Index | 29 | — |
| EBR-bench | 92.4 | — |
| GDP.pdf | 29.8 | — |
| MMMU-Pro | 89.5 | — |
| AA-AnalystAgent | 62.5 | — |
| BenchCAD 2.1 Core | 82.5 | — |
| EMB (Excel Modeling) | 76.0 | — |
| Harvey LAB-AA v1.1 (All-pass) | 15.0 | — |
| Harvey's Legal Agent Benchmark | 16.3 | — |
| ARC-AGI-1 | 98.5 | — |
| ARC-AGI-2 | 95.8 | — |
| ARC-AGI-3 | 100.0 | — |
| SimpleBench | 77.3 | — |
| SimpleQA Verified | 78.5 | — |
| BioMysteryBench | 80.7 | — |
| HealthBench Professional | 66.3 | — |
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix · 5% steps
Blended price = effective input rate at 90% cache hits × 80% + output rate × 20%
GPT-6.2 Astra costs10,379%morethan Mercury 3 per million tokens — 104.8× the price.Using listed rates vs standard rates at this token mix.
GPT-6.2 Astra: Standard pricing. context_length <= 272000 Official OpenAI per-million-token pricing. Full request uses the applicable input-token band. context_length > 272000
Mercury 3: Standard pricing. Inception Labs official docs/get-started/models.md verified 2026-11-13 (50% promo pricing prior to 2026-12-28 wind-down); Mercury 3 with 50% cache read promo
| Price / 1M tokens | GPT-6.2 Astra | Mercury 3 | Ratio |
|---|---|---|---|
| Input | $8.00 | $0.10 | 80.0× |
| Output | $40.00 | $0.35 | 114× |
| Cached input | $0.40 | $0.01 | 40.0× |
| Cache write | $10.00 | — | — |
| Blended | $8.93 | $0.085 | 105× |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
GPT-6.2 Astra relative to Mercury 3
No matching recorded settings for a cost summary.
Summary: 0 of 5 shared benchmarks have matching recorded settings and sources. The average is a geometric mean of 0 positive-cost ratios, one equal weight per benchmark. This benchmark set can include related tests and indexes; it is not a workload cost estimate.
Settings or sources differ
Settings or sources differ
Settings or sources differ
Settings or sources differ
Settings or sources differ
Settings or sources differ
Settings or sources differ
Settings or sources differ
Settings or sources differ
Settings or sources differ
Settings or sources differ or are not recorded; interpret the comparison with care.
GPT-6.2 Astra
Mercury 3
Settings or sources differ or are not recorded; interpret the comparison with care.
GPT-6.2 Astra
Mercury 3
Settings or sources differ or are not recorded; interpret the comparison with care.
GPT-6.2 Astra
Mercury 3
Settings or sources differ or are not recorded; interpret the comparison with care.
GPT-6.2 Astra
Mercury 3
Settings or sources differ or are not recorded; interpret the comparison with care.
GPT-6.2 Astra
Mercury 3
| Benchmark | GPT-6.2 Astra | Mercury 3 |
|---|---|---|
| AA Intelligence Index v5.0.1 | 2 more settings | Not reported |
| Agent Arena Index | $3.2842 / task Median · Agent Arena Index rating (index points): 1291 Max (With Tools + Internet) | Not reported |
| ARC-AGI-1 | 2 more settings | Not reported |
| ARC-AGI-2 | 2 more settings | Not reported |
| BenchCAD 2.1 Core | $2.89 / task Average · Core mean geometry/circuit score (0–100): 74.6 Thinking High (With Tools) $0.98 / task Average · Core mean geometry/circuit score (0–100): 42.6 Thinking Low (With Tools) 2 more settings$1.94 / task Average · Core mean geometry/circuit score (0–100): 68.5 Thinking Medium (With Tools) $3.36 / task Average · Core mean geometry/circuit score (0–100): 77.6 Extra-High (With Tools) | Not reported |
| BioMysteryBench | Not reported | |
| EMB (Excel Modeling) | Not reported | |
| FrontierCode 2.0 | $3.952 / task Average · Weighted rubric score: 43.6 Thinking High (With Tools + Internet) 2 more settings$2.469 / task Average · Weighted rubric score: 42.2 Thinking Medium (With Tools + Internet) | Not reported |
| Harvey's Legal Agent Benchmark | Not reported | |
| MysteryMechanism | Not reported | |
| Terminal-Bench-Science 0.1 | Not reported |
| Attribute | GPT-6.2 Astra | Mercury 3 |
|---|---|---|
| Availability | Available | Available |
| Provider | OpenAI | Inception Labs |
| Model type | Reasoning model | Reasoning model |
| First release | November 17, 2026 | October 19, 2026 |
| Context window | 1.05M tokens | 1M tokens |
| Max output | 128K tokens | 128K tokens |
| Architecture | Undisclosed | Undisclosed |
| Documented capabilities | Reasoning, Coding, Function calling, Tool use, Structured output, Computer use, Web search, Long context | Reasoning, Tool use, Function calling, Structured output, Long context, Coding |
| Knowledge cutoff | 2026-05-18 | — |
| Input modalities | Text, Image | Text |
| Output modalities | Text | Text |
| Weights | Closed / API only | Closed / API only |
| Commercial use | Proprietary | Proprietary |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used GPT-6.2 Astra or Mercury 3? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.