Google launches Gemini 4 Hydrogen, replacing Flash series

Google released Gemini 4 Hydrogen on November 13, 2026, a faster, cheaper model targeting coding, document processing, and multi-step agents. It runs 2.5x faster than Gemini 3.8 Flash, priced at $1.50/$7.50 per million input/output tokens. It scores 57% on Terminal-Bench 4.1, behind Claude Sonnet 5.5's 63.9%. Google's Antigravity Orchestrator pairs it with Gemini 4 Carbon, running 19% faster using 24% less quota.

Gemini 4 Hydrogen launch

• ▪Gemini 4 Hydrogen is designed for writing code, working through documents, and running multi-step AI agents
• ▪Google released Gemini 4 Hydrogen
• ▪Gemini 4 Hydrogen is rolling out globally via the Gemini API, AI Studio, Gemini for business, the Gemini app, and Google Search
• ▪Google Cloud lists Gemini 4 Hydrogen as generally available under the model name gemini-4-hydrogen

Performance benchmarks

• ▪Box CEO Aaron Levie stated that Gemini 4 Hydrogen took less than half as long and used about 12% fewer tokens on a typical task compared to Gemini 3.8 Flash
• ▪In Box's Complex Work Eval early-access test, accuracy improved from 74% with Gemini 3.8 Flash to 79% with Gemini 4 Hydrogen
• ▪On Google's Terminal-Bench 4.1 evaluation, Gemini 4 Hydrogen completes 57% of tasks, compared with 18.2% for Gemini 3.8 Flash
• ▪Google says Gemini 4 Hydrogen completes tasks 2.5 times as fast as Gemini 3.8 Flash at Hydrogen's default reasoning setting
• ▪In Artificial Analysis's Intelligence Index testing, Gemini 4 Hydrogen at its default reasoning setting scored 51, up from 31 for Gemini 3.8 Flash at high reasoning

Pricing comparison

• ▪Google lists no introductory promotion for Gemini 4 Hydrogen's standard tier
• ▪Gemini 4 Hydrogen's Contributor tier is invite-only and offers standard-tier speed and reliability in exchange for letting Google use data from participating requests to improve its products and train models
• ▪Gemini 3.8 Flash's introductory pricing of $0.75 per million input tokens and $3.75 per million output tokens runs through December 31, 2026, after which rates rise to $1.50 input and $7.50 output — matching Gemini 4 Hydrogen's standard pricing
• ▪Gemini 4 Hydrogen's Batch, Flex, and Contributor tiers each cost half the standard rates
• ▪Google's standard paid Gemini API rate for Gemini 4 Hydrogen is $1.50 per million uncached input tokens, $0.075 per million cached input tokens, and $7.50 per million output tokens
• ▪Gemini 4 Hydrogen costs 15 times as much per uncached input and output token as the lowest standard rates for Anthropic's Claude Haiku 5.5, OpenAI's GPT-6 Luna, and Meta's Muse Ember

Migration technical changes

• ▪Gemini 4 Hydrogen accepts text, images, audio, and video, and Google Cloud can ingest Office and OpenDocument files directly
• ▪Gemini 4 Hydrogen defaults to high reasoning, up from medium on Gemini 3.8 Flash, and adds an xhigh reasoning option
• ▪Explicit caching is unavailable for Gemini 4 Hydrogen on Google Cloud at launch
• ▪Gemini 4 Hydrogen has a roughly 1 million-token context window shared by input, output, reasoning tokens, and tool calls
• ▪Gemini 4 Hydrogen rejects the old thinking_budget parameter and the MINIMAL reasoning setting used by Gemini 3.8 Flash
• ▪On Google Cloud, Gemini 4 Hydrogen's automatic caching starts at 4,096 tokens, down from 6,144 tokens on Gemini 3.8 Flash

Chinese open-weight rivals

• ▪Kimi K3.1 Pro's weights come with Moonshot's own license, and multiple providers host the model
• ▪Moonshot's Kimi K3.1 Pro scores 46 on the Artificial Analysis Intelligence Index at an estimated $3.68 per benchmark task through Kimi's API
• ▪DeepSeek V4.1 Flash 1017 offers text and image input, a million-token context window, and MIT-licensed weights
• ▪Artificial Analysis estimates Xiaomi's hosted MiMo-V2.6-Flash model costs $0.13 per benchmark task, versus $1.87 for Gemini 4 Hydrogen at high reasoning, with a 24-point gap in the composite Intelligence Index score
• ▪DeepSeek's hosted API for DeepSeek V4.1 Flash 1017 drops to $0.15 per million input tokens and $0.60 per million output tokens off-peak
• ▪Xiaomi's MiMo-V2.6-Flash offers MIT-licensed weights with audio and video support documented for the downloadable model

Antigravity Orchestrator integration

• ▪Google's Antigravity Orchestrator can switch between Gemini 4 Carbon and Gemini 4 Hydrogen within the same session
• ▪Google reports that the Antigravity Orchestrator combining Gemini 4 Carbon and Gemini 4 Hydrogen came close to Gemini 4 Carbon's success rate on Terminal-Bench 4.0 while finishing 19% faster and using 24% less quota