Gemini 4 model guide, pricing and effort-level cost benchmarks released
Official guides and hands-on details for the three Gemini 4 models have surfaced together. According to the Gemini 4 model guide shared by @patloeber, Carbon, Argon, and Hydrogen share a unified 1M context window, a long-decode architecture, and built-in tools; Carbon is the strongest frontier model, spending more thinking tokens per turn to verify hypotheses and edge cases, making it suited for architecture design and tricky debugging; Hydrogen is priced at $1.50/$7.50 per million input/output tokens. Inference platform Runware announced that Carbon and Hydrogen are now live on its OpenAI-compatible API, confirming the 1M context and pricing.
Confirmed
- The three Gemini 4 models share a 1M context window, long-decode architecture, and built-in tools
- Gemini 4 Hydrogen is priced at $1.50/$7.50 per million input/output tokens
- Carbon and Hydrogen are live on Runware's OpenAI-compatible API
- Paweł Huryn's subscription-tier comparison table (corrected based on reader feedback, using US prices as of November 2026): student/trial Pro reportedly gets only Flash; paid Pro tops out at Argon; using Carbon requires an Ultra subscription
- Huryn published real-world cost reports for Hydrogen, Argon, and Carbon across all effort levels, presented as a filterable online tool with reference models for comparison
Not Confirmed
- The claim that student/trial Pro gets only Flash comes from reader feedback and is marked by Huryn as "reported"
Why It Matters
- The guide clarifies each model's positioning and recommended combinations, giving an official basis for model selection; the real-world cost data lets users weigh per-task cost by effort level rather than looking only at unit price
2026-11-14 ~ 2026-11-14 · 7 related posts
- Episode 1: Rumored Gemini 4 Carbon Reportedly Beats Higher-Tier Argon(2026-11-12, 4 posts)
- Episode 2: Google Launches Gemini 4 Hydrogen Across Apps, Search and APIs(2026-11-13, 81 posts)
- Episode 3: Gemini 4 Hydrogen benchmarks leak: near-flagship intelligence at a fraction of the cost(2026-11-13, 26 posts)
- Episode 4: Gemini 4 Hydrogen Ranks 5th on Agent Arena as Fastest Model(2026-11-13, 4 posts)
- Episode 5: Researchers slam Gemini's chemical-element model naming as confusing(2026-11-13, 2 posts)
- Episode 6: Google Help Page Leaks Unreleased Gemini Models Hydrogen and Argon(2026-11-13, 2 posts)
- Episode 7: Gemini 4 model guide, pricing and effort-level cost benchmarks released(2026-11-14, 7 posts)
- Episode 8: Google Unleashes a Week of Releases: Twin TPUs, Gemini 4 Variants, Robotics 2.5(2026-11-14, 2 posts)
Primary sources
- Pawel Huryn charts cost-per-task across models and effort levels — PawelHuryn · 2026-11-14
- [source] Gemini 4 Hydrogen, Argon, Carbon Tested Across All Effort Levels with Live Report — PawelHuryn · 2026-11-14
- Gemini 4 Carbon and Hydrogen hit Runware with 1M context and new pricing — aziz4ai · 2026-11-14
- [source] Gemini plan-to-model map corrected: paid Pro caps at Argon, Carbon needs Ultra or API — PawelHuryn · 2026-11-14
- Chart: Which Gemini 4 model each Google subscription tier actually gets — PawelHuryn · 2026-11-14
- Gemini 4 guide: Carbon is the frontier model, Hydrogen costs $1.50/1M input — patloeber · 2026-11-14
- [source] Full Gemini 4 model guide surfaces with selection and pairing advice — patloeber · 2026-11-14