Gemini 3.8 Flash vs Gemini 4 Argon
Different efforts: medium · high · Compare all at high
MMMU-Pro (vision) MMMU-Pro (vision): College-level questions that require understanding images: charts, diagrams, photos
Gemini 3.8 Flash84.2
Gemini 4 Argon—
Chartography Chartography: Expert chart reasoning over professional scientific, medical, financial and engineering charts
Gemini 3.8 Flash · high40.9
Gemini 4 Argon71.6
GDP.pdf · AA all-pass GDP.pdf · AA all-pass: Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes
Gemini 3.8 Flash22.8
Gemini 4 Argon21.8
Arena Vision · overall Arena Vision · overall: Human pairwise preferences in the stated vision vote pool
Gemini 3.8 Flash · high1,315.4
Gemini 4 Argon—
| Benchmark | Gemini 3.8 Flash | Gemini 4 Argon |
|---|---|---|
| MMMU-Pro (vision) | 84.2 | — |
| Chartography | 40.9 · high | 71.6 |
| GDP.pdf · AA all-pass | 22.8 | 21.8 |
| Arena Vision · overall | 1,315.4 · high | — |
Specs
| Spec | Gemini 3.8 Flash | Gemini 4 Argon |
|---|---|---|
| Input, per 1M | !$0.75 | !$2 |
| Output, per 1M | !$3.75 | !$10 |
| Cost per task? | $0.93 | $1.99 |
| Speed, tok/s | — | 84 |
| Context | 1M | 1M |
| Weights | Closed | Closed |
| Released | Sep 2026 | Oct 2026 |
| See images | ✓ | ✓ |
| Read PDFs | ✓ | ✓ |
| Hear audio | ✓ | ✓ |
| Watch video | ✓ | ✓ |
| Call tools | ✓ | ✓ |
| Force a tool call | Partial | Partial |
| Return strict JSON | ✓ | ✓ |
| Search the web | ✓ | ✓ |
| Run code | ✓ | ✓ |
| Use a computer | ? | ? |
What breaks when you switch
Gemini 3.8 Flash → Gemini 4 Argon
Synchronous output stops at 65,536›
With background false, output limits above 65,536 tokens return HTTP 400; stay at or below that limit.
Sampling parameters are ignored›
temperature, top_p and top_k are ignored; remove them from requests.
No explicit caching›
Explicit caching is unavailable; omit explicit cache requests.
Gemini 4 Argon → Gemini 3.8 Flash
Sampling parameters are ignored›
temperature, top_p and top_k are ignored now and will cause HTTP 400 in future generations; remove them from requests.
No minimal thinking level›
The minimal thinking level returns an error; avoid minimal.