Editorial Team・Oct 11, 2026・6 mins read
Gemini 3.8 Flash is Google's newest Flash model, released on September 2, 2026 — the third Flash release in about six weeks. Google calls it its best reasoning and coding model yet at the same speed and low cost as Gemini 3.7 Flash, and on several benchmarks it lands within a point of far more expensive models. If you use Gemini 3.5 Flash today, it is a few versions ahead.
This Gemini 3.8 Flash review covers what changed, what the benchmark numbers show, what it costs now and after its price doubles in January, and where its limits sit. It also covers access: Gemini 3.8 Flash is not yet integrated into Chat Smith, but several other Gemini models already are.
Gemini 3.8 Flash is the fast, lower-cost workhorse of Google's Gemini family. According to Google's model card, it is built on Gemini 3.7 Flash and tuned to work harder on difficult tasks: taking extra reasoning steps, calling tools repeatedly, and checking its work before answering.
It accepts text, images, video, audio, and PDFs, returns text, and supports a 1,048,576-token context window with up to 65,536 tokens of output. You can set its thinking level to low, medium (the default), or high, trading speed for depth. Its knowledge cutoff is March 2026, and computer use is available in preview.
Google has moved quickly through the Flash line this summer: 3.6 Flash in July, covered in our Gemini 3.6 Flash review, then 3.7 Flash in August and 3.8 Flash in September. Above the Flash tier sits Gemini 4 Argon, Google's new frontier model, which is still in limited release.
Four areas show the biggest gains over Gemini 3.7 Flash.
On DeepSWE v1.1, a test of real-world software engineering, Gemini 3.8 Flash scores 73.7%, up from 65.3% for Gemini 3.7 Flash. That puts a Flash model within 0.3 points of Claude Opus 5 (74.0%) and ahead of GPT-5.6 Sol (72.7%). On Terminal-bench 2.1 it reaches 89.4%, slightly ahead of both.
Its strongest wins in Google's table are in professional work. It posts the top score on Vals Finance Agent v2 at 61.4% and on Harvey's Legal Agent Benchmark at 10.0%, ahead of Claude Opus 5 and GPT-5.6 Sol on both. For jobs like pulling figures out of reports or running long filings through an AI PDF summarizer, this is where the upgrade shows.
Gemini 3.8 Flash is especially strong with visual material. On LVBench, a long-video test, it scores 87.8% in agentic mode, more than 12 points ahead of Claude Opus 5, and on CharXiv Reasoning, which tests chart understanding, it reaches 86.2%. That makes it a good fit for tasks like asking an AI to summarize a YouTube video or explain a dense chart.
On HLE-Verified, a set of expert-level questions, it scores 54.9%, narrowly ahead of GPT-5.6 Sol (54.5%) and Claude Opus 5 (54.4%). The margins are small, but they are notable for a model in the Flash price tier. If you regularly work through hard problems with an AI math solver, that extra reasoning depth is useful.
The table below uses Google's launch figures, comparing Gemini 3.8 Flash with its predecessor and with Claude Opus 5, the strongest rival in Google's table. Prices are API rates per million tokens.
| Gemini 3.8 Flash | Gemini 3.7 Flash | Claude Opus 5 | |
|---|---|---|---|
| Input / output price | $0.75 / $3.75 (intro) | $0.75 / $3.75 (intro) | $5 / $25 |
| DeepSWE v1.1 | 73.7% | 65.3% | 74.0% |
| Terminal-bench 4.0 | 19.1% | 11.2% | 51.8% |
| OSWorld-2.0 (computer use) | 59.0% | 50.6% | 75.4% |
| Vals Finance Agent v2 | 61.4% | 59.0% | 58.6% |
| HLE-Verified | 54.9% | 53.6% | 54.4% |
| LVBench (long video) | 87.8% | 85.4% | 75.4% |
| Available in Chat Smith | Not yet | Not yet | — |
The pattern is clear. Gemini 3.8 Flash beats its predecessor on every row and comes close to or beats Claude Opus 5 on coding, finance, reasoning, and video, at roughly a seventh of its input price. Opus 5 still leads by a wide margin on long, multi-step agent tasks and computer use.
Gemini 3.8 Flash costs $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026. On January 1, 2027, those rates double to $1.50 and $7.50. Batch and Flex requests cost half, cached input is discounted by 90%, and there is no surcharge for long prompts. Thinking tokens are billed as output, so harder tasks can cost more than the per-token price suggests.
Gemini 3.8 Flash works harder by design, which can make it feel slow. Google warns that it may use more tokens at higher effort levels, and its model card mentions occasional slowness or timeouts. For quick tasks, set the thinking level to low.
Long agent runs and computer use remain weak spots. On Terminal-bench 4.0 it scores 19.1% against 51.8% for Claude Opus 5, and on OSWorld-2.0 it trails 59.0% to 75.4%. If your work involves long, multi-step automation, test alternatives such as Claude Sonnet 5 first.
Pricing will also change. The introductory rate ends on December 31, 2026, so it is safer to budget at the 2027 price of $1.50 and $7.50. The benchmark figures are Google's own, independent results are still settling, and your own workload is the real test.
Gemini 3.8 Flash is not integrated into Chat Smith yet, but several Gemini models already are. Pick one from the model list and start chatting on Android, iOS, or in the browser, with no Google AI subscription needed.
Gemini 3.5 Flash is the closest option for everyday work: fast answers, writing, research, and summaries at Flash-tier speed. For very quick, lightweight tasks, Gemini 3.1 Flash Lite is even lighter, and for longer documents or more careful analysis, Gemini 2.5 Pro offers more depth. Gemini 3.8 Flash's biggest gains show up mainly in coding agents and specialist finance or legal work.
Because Chat Smith also includes models from OpenAI, Anthropic, xAI, and DeepSeek, you can put the same prompt to a Gemini model and its rivals and keep the answer that holds up. Chat Smith Pro adds higher usage limits and more premium models for $9.99 per month or $39.99 per year.
Gemini 3.8 Flash shows how far the Flash tier has come: near-Opus coding scores, top results on finance, legal, video, and expert reasoning tests, a 1M-token context window, and a price that stays low even after it doubles in 2027. Its weak spots are long agent runs, computer use, and slower answers on hard tasks.
Until it reaches more apps, Google's current Gemini models remain an easy way in. You can try them now in Chat Smith, an AI chat app that puts Gemini next to the latest models from OpenAI, Anthropic, xAI, and DeepSeek.
Gemini 3.8 Flash is Google's newest Flash model, released on September 2, 2026, built on Gemini 3.7 Flash with stronger reasoning and coding, a 1M-token context window, and support for text, image, video, audio, and PDF input. Chat Smith has not integrated Gemini 3.8 Flash yet, but you can use other Gemini models there alongside leading AI models from other labs in one app.
The Chat Smith Editorial Team brings together AI enthusiasts, researchers, and creators to make AI accessible and practical. We share trends, tool reviews, and actionable guides that help individuals and businesses stay informed, boost productivity, and get more from AI.
