← All ComparisonsGemini 3.7 Flash vs Gemini 3.8 Flash
| Gemini 3.7 Flash | Gemini 3.8 Flash |
|---|
| Pricing |
|---|
| Input / output per 1M* | $0.75 / $3.75 | $0.75 / $3.75 |
|---|
| From Jan 2027 | $1.50 / $7.50 | $1.50 / $7.50 |
|---|
Introductory rate through Dec 31, 2026.
Identical cards - 3.8 can still cost more per task because it spends more tokens at High effort.
| Gemini 3.7 Flash | Gemini 3.8 Flash |
|---|
| Coding |
|---|
| DeepSWE v1.1 | 65.3% | 73.7% |
|---|
| Terminal-Bench 4.0 | 11.2% | 19.1% |
|---|
Both rows move strongly, but 19.1% still trails Sol's 37.3% on the broader agent test.
| Gemini 3.7 Flash | Gemini 3.8 Flash |
|---|
| Reasoning and science |
|---|
| BioMysteryBench (hard set) | 43.5% | 56.5% |
|---|
The largest single jump in the table at 13 points; HLE-Verified, GDP.PDF and Harvey legal each rise about a point.
| Gemini 3.7 Flash | Gemini 3.8 Flash |
|---|
| Agentic work |
|---|
| OSWorld 2.0 | 50.6% | 59.0% |
|---|
Up 8.4 points on computer use, though the two figures come from different task setups.
| Gemini 3.7 Flash | Gemini 3.8 Flash |
|---|
| Overall intelligence |
|---|
| High effort | 56 | 59 |
|---|
| Medium effort | 53 | 57 |
|---|
| Low effort | 51 | 52 |
|---|
Gains concentrate at higher effort - Medium gains four points, Low gains one.
| Gemini 3.7 Flash | Gemini 3.8 Flash |
|---|
| Context and availability |
|---|
| Context window | 1M tokens | 1M tokens |
|---|
| Cyber variant | None | Flash Cyber (Fairwind) |
|---|
Same window; the only structural addition is the restricted Cyber tier.