The first entry is always the pin's source. Overall confidence is a weighted average of how firmly each reference supports the start and end times used above; a reference counts half as much for every 180 days older than the newest.
Google's[2][9] own post, dated Sep 02, 2026, says "today we're introducing Gemini 3.8", and CNBC[6] reports the company "on Wednesday launched Gemini 3.8 Flash".
Google[9] Cloud's official developer guide gives the model's specifications: model ID gemini-3.8-flash, GA launch stage, 1,048,576-token context window, 65,536 max output tokens, and LOW/MEDIUM/HIGH thinking levels.
Google's[2][9] follow-up post of September 15, 2026 extending the 3.8 generation to its live voice models, Gemini 3.8 Live and 3.8 Live Extended Thinking.
Google's[2][9] same-day post launching the Fairwind Program, which gives governments, critical infrastructure operators and core technology platforms access to Gemini 3.8 Flash Cyber with the CodeMender harness.
Independent same-day report that Google[2][9] is rolling out Gemini 3.8 Flash today, with its knowledge cutoff, introductory pricing until December 31, availability and the Flash Cyber model.
Something missing or wrong? Say it in your own words: a link that backs this pin up, a different start or end date and why, or a fact it lacks or gets wrong. The AI checks it against this pin's sources, searches for better ones, and adds any page that backs you up. The pin's own sources still count most. A picture that shows something else, or shows it badly, is looked at too, and moved down or replaced.
| median Gemini Apps text prompt, all models, May 2025 |
| Carbon per prompt | 0.03 gCO2e | same prompt, same basis |
|---|
| Water per prompt | 0.26 mL | about five drops, same prompt |
|---|
| Efficiency gain | 33x energy, 44x carbon | over the 12 months to May 2025, Google[2][9] claim not independently verified |
|---|
| Gemini 3.8 Flash alone | Not published | the launch post gives no energy, carbon or water figure |
|---|
Google's only first-party per-prompt figure; it covers text prompts, not heavy agentic or multimodal use, or training. · source
| Measure | Figure | Basis |
|---|---|---|
| Gemini app users | 950 million | monthly active, Google[2][9]-wide, 22 Jul 2026 |
| API throughput | 22 billion tokens/min | up from 16 billion a quarter earlier |
| Developers | 9 million+ | building each month across Google's[2][9] APIs and products |
| Gemini Enterprise | ~90% of Fortune 100 | Google[2][9]-wide adoption |
| Fairwind partners | 650+ | Flash Cyber, trusted defenders only |
| 3.8 Flash surfaces | 7 | app, AI Mode, Sheets, Antigravity, AI Studio, API, Enterprise |
Usage figures predate the 2 Sep 2026 launch and are not split by model, so they are not 3.8 Flash alone. · source
| Benchmark | Gemini 3.8 Flash | Gemini 3.7 Flash | Claude Opus 5 | Claude Sonnet 5 | GPT-5.6 Sol | GPT-5.6 Terra |
|---|---|---|---|---|---|---|
| Input price ($/1M tokens, no caching) | $0.75 | $0.75 | $5.00 | $2.00 | $4.00 | $2.00 |
| Output price ($/1M tokens) | $3.75 | $3.75 | $25.00 | $10.00 | $20.00 | $12.00 |
| DeepSWE v1.1 (Long-horizon software engineering) | 73.7% | 65.3% | 74.0% | 53.8% | 72.7% | 69.6% |
| GDPVal-AA v2 (Knowledge work, Elo) | 1545 | 1482 | 1824 | 1584 | 1710 | 1528 |
| Vals Finance Agent v2 (Financial analyst tasks) | 61.4% | 59.0% | 58.6% | 53.9% | 53.8% | 54.4% |
| Harvey's Legal Agent Benchmark (Complex legal workflows, All pass rate) | 10.0% | 8.8% | 6.7% | 5.0% | 2.5% | 0.8% |
| Terminal-bench 2.1 (Agentic terminal coding) | 89.4% | 85.8% | 89.1% | 80.4% | 88.8% | 87.4% |
| Terminal-bench 4.0 (General agent capabilities) | 19.1% | 11.2% | 51.8% | 12.4% | 37.3% | 23.6% |
| GDP.PDF (Expert PDF document comprehension, All pass rate) | 35.0% | 34.0% | 37.0% | 28.0% | 40.0% | 29.0% |
| CharXiv Reasoning (Information synthesis from complex charts, No tools) | 86.2% | 84.5% | 83.7% | 70.1% | 85.8% | 85.9% |
| LVBench (Long video understanding) | 87.8% | 85.4% | 75.4% | 68.5% | 82.1% | 78.9% |
| LVBench (Long video understanding) | 87.1% | 85.4% | 75.4% | 68.5% | 82.1% | 78.9% |
| HLE-Verified (Multidisciplinary expert reasoning) | 54.9% | 53.6% | 54.4% | 31.0% | 54.5% | 51.1% |
| OSWorld-2.0 (Agentic computer use, Partial scorebatch tool enabled) | 59.0% | 50.6% | 75.4% | 42.6% | 62.6% | 50.2% |
| BioMysteryBench (Bioinformatics research workflows, Human Solvable) | 88.8% | 87.1% | 90.1% | 87.5% | 79.5% | 83.8% |
| BioMysteryBench (Bioinformatics research workflows, Human Difficult) | 56.5% | 43.5% | 49.4% | 34.1% | 44.7% | 49.4% |
| LABBench2 (Biology real-world research tasks) | 86.2% | 82.1% | 84.2% | 80.1% | 82.1% | 81.2% |