Skip to content
Main Site News Console

llm-gemini 0.35 Released

· Simon Willison Translated
技巧LLM

Releases: llm-gemini 0.35—an LLM plugin for accessing the Google Gemini family of models

I added support for the brand-new Gemini 4 Argon model to llm-gemini. It became available today via the Gemini API and to subscribers on the $100/month plan.

Here are low, medium, and high-quality versions of a pelican. Unsurprisingly, they all look fantastic: Gemini has consistently excelled at generating great SVGs, and as far as I know, they are the only lab with staff actively dedicated to the SVG problem space (having previously shown off their pelican generations as well).

Here is the “high” quality illustration, which cost me 17.166 cents:

Figure 1: Stylish, structurally accurate bicycle frame. Fish in the basket. Snazzy scarf. Feet on pedals on either side of the frame. Wheel spokes, small hat, excellent beak and throat pouch, slightly majestic expression.

Its pricing is fascinating. At launch, Argon is priced at $2 per million input tokens and $10 per million output tokens—matching GPT-6 Sol, half the price of Claude Opus 5.5, and a fifth of Claude Fable 5.1 and GPT-6 Astra.

Come December 1st, however, Argon’s price will double—putting it on par with Opus 5.5.

Comparing list prices in isolation doesn’t mean much these days, because what really matters is how many tokens a model requires to accomplish a goal within an agentic harness.

Artificial Analysis provides a very useful signal here by publishing the total cost of running their benchmarks across different models. Argon looks extremely compelling here—running the benchmark costs $2,407, close to GLM-5.3, and significantly cheaper than the latest models from OpenAI, Anthropic, and Meta.

Even after its price doubles, it will still remain slightly cheaper than GPT-6 Astra.

Figure 2: Screenshot of a stacked bar chart from Artificial Analysis comparing the dollar cost of 14 AI models, ordered from lowest to highest, bars colored by provider, each bar featuring the provider's logo underneath, with model names angled at the bottom. Top right logo is Artificial Analysis. Tooltip hovering over the Gemini 4 Argon bar reads: Gemini 4 Argon (High), $2,407, Output $313, Reasoning $815, Cache Write $897, Cache Read $264, Non-Cached Input $118, click for more details. Bars from left to right show total cost above each bar, with some segment values printed inside: Qwen3.8 27B (xhigh), $1,336, Alibaba logo; Mistral Large 4 Preview, $1,607, Mistral logo; Gemini 3.8 Flash (high), $1,623, Google logo; Muse Spark 1.3 (max), $2,000, segment $769, Meta logo; Gemini 4 Argon (high), $2,407, segments $815 and $897, Google logo; GLM-5.3 (max), $2,503, segment $790, Z.ai logo; Kimi K3 (max), $3,658, Kimi logo; Qwen3.8 Max (0902), $4,935, segments $893 and $1,441, Alibaba logo; Grok 4.7 (xhigh), $4,967, segments $1,861 and $1,618, xAI logo; GPT-6 Astra (max), $5,324, segments $778 and $978, OpenAI logo; Muse Wonder (max), $5,733, segment $1,748, Meta logo; Claude Sonnet 5.5 (max with fallback), $7,260, segments $1,216 and $1,815, Anthropic logo; Claude Opus 5.5 (max with fallback), $8,708, segment $921, Anthropic logo; Claude Fable 5.1 (max with fallback), $13,129, segment $7,219, Anthropic logo.

(I edited the chart using ChatGPT! The data is fully verified—I just needed to display the Gemini 4 Argon tooltip without it overlapping other numbers.)

A few more examples using Gemini 4 Argon: here’s an SVG of “an armadillo jaywalking on Mars wearing fishnet stockings”; here’s a Fabergé egg it generated for me in Blender; and DeepMind employee @fofrAI shared a video drawing a pelican on a bicycle using Fourier epicycles. What a fun new toy!