Compare / head-to-head
Claude Opus 4.7vsGemini 3.8 Flash
Gemini 3.8 Flash leads 3 of 5 shared benchmarks. Gemini 3.8 Flash is 6.7x cheaper per token. Gemini 3.8 Flash has the larger context window (1.0M tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
2 – 3
Gemini 3.8 Flash leads
Cheaper per token
Gemini 3.8 Flash
6.7x cheaper, input + output
Larger context
Gemini 3.8 Flash
1.0M tokens
Provider uptime (30d)
100% · 100%
Anthropic · Google
Claude Opus 4.7
Anthropic · released 2026-04-16
textvision
Gemini 3.8 Flash
Google · released 2026-09-02
textvisionaudio
Quality
Benchmark matrix
5 shared · 12 only Claude Opus 4.7 · 10 only Gemini 3.8 Flash| Benchmark | Claude Opus 4.7 | Gemini 3.8 Flash | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 5 | ||||
| GPQA Diamond | 94.2% ↗ | 95.3% ↗ | -1.1 pt | Gemini 3.8 Flash |
| FrontierMath Tier 4 v2 (Epoch AI run) | 31.7% ↗epoch | 22% ↗epoch | +9.7 pt | Claude Opus 4.7 |
| FrontierMath Tiers 1-3 v2 (Epoch AI run) | 70.2% ↗epoch | 68.4% ↗epoch | +1.8 pt | Claude Opus 4.7 |
| OTIS Mock AIME 2024-2025 (Epoch AI run) | 97.8% ↗epoch | 98.9% ↗epoch | -1.1 pt | Gemini 3.8 Flash |
| SimpleQA Verified | 51.7% ↗epoch | 69.7% ↗epoch | -18 pt | Gemini 3.8 Flash |
| Only Claude Opus 4.7 reports · 12 | ||||
| SWE-bench Verified | 87.6% ↗ | not reported | — | — |
| AIME 2026 | 95.8% ↗matharena ⚠ | not reported | — | — |
| Humanity's Last Exam (no tools) | 46.9% ↗ | not reported | — | — |
| LMArena Elo | 1483.4 ↗ | not reported | — | — |
| BrowseComp | 79.8% ↗ | not reported | — | — |
| Humanity's Last Exam (with tools) | 54.7% ↗ | not reported | — | — |
| OSWorld-Verified | 82.8% ↗ | not reported | — | — |
| SWE-bench Multilingual | 80.5% ↗ | not reported | — | — |
| SWE-bench Multimodal | 34.5% ↗ | not reported | — | — |
| SWE-Bench Pro | 64.3% ↗ | not reported | — | — |
| SWE-bench Verified (Epoch AI run) | 83.5% ↗epoch | not reported | — | — |
| Terminal-Bench 2.1 | 66.1% ↗ | not reported | — | — |
| Only Gemini 3.8 Flash reports · 10 | ||||
| CharXiv Reasoning (no tools) | not reported | 86.2% ↗ | — | — |
| DeepSWE v1.1 | not reported | 73.7% ↗ | — | — |
| GDPVal-AA v2 | not reported | 1545 ↗ | — | — |
| Harvey's Legal Agent Benchmark | not reported | 10% ↗ | — | — |
| HLE-Verified | not reported | 54.9% ↗ | — | — |
| LABBench2 | not reported | 86.2% ↗ | — | — |
| OSWorld-2.0 | not reported | 59% ↗ | — | — |
| Terminal-bench 2.1 | not reported | 89.4% ↗ | — | — |
| Terminal-bench 4.0 | not reported | 19.1% ↗ | — | — |
| Vals Finance Agent v2 | not reported | 61.4% ↗ | — | — |
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Claude Opus 4.7 minus Gemini 3.8 Flash in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing
Side by side
Official pricing pages and model cards| Spec | Claude Opus 4.7 | Gemini 3.8 Flash | Edge |
|---|---|---|---|
| Context window | 1M tokens | 1.0M tokens | Gemini 3.8 Flash |
| Max output | 128K tokens | 66K tokens | Claude Opus 4.7 |
| Input price / 1M | $5 ↗ | $0.75 ↗ | Gemini 3.8 Flash |
| Output price / 1M | $25 ↗ | $3.75 ↗ | Gemini 3.8 Flash |
| Cached input / 1M | $0.5 ↗ | $0.075 ↗ | Gemini 3.8 Flash |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | $30 | $4.5 | Gemini 3.8 Flash |
| Modalities | text · vision | text · vision · audio | Gemini 3.8 Flash |
| Released | 2026-04-16 | 2026-09-02 | — |
| Cited benchmark scores | 18 | 16 | — |
Reliability
Provider status
Last 60 days · refreshed daily on this page · live on /statusMore matchups
Claude Opus 4.7 vs …
Models sharing the most benchmarksMore matchups
Gemini 3.8 Flash vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Claude Opus 4.7 and Gemini 3.8 Flash on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.