llm-gemini 0.35

What it claims

llm-gemini 0.35 adds support for Gemini 4 Argon, released on 13 October 2026 through the Gemini API and to customers of Google's $100/month subscription plan. The page places Argon's launch price of $2/M input and $10/M output against named models:

The price doubles on December 1st, which makes Argon the same price as Opus 5.5. If you are choosing a model on price, the launch window runs only until then.

Evidence

• Agentic cost proxy. Artificial Analysis publishes the total cost of running its benchmark with each model. For Argon that cost is $2,407. This is close to GLM-5.3 and "significantly cheaper" than the latest models from OpenAI, Anthropic and Meta. The page gives no per-model figures for those competitors.
• After the December price rise. The author says Argon will "still be a little cheaper than GPT-6 Astra." In context this seems to refer to the benchmark run cost, so after December 1st the cost advantage shrinks to a narrow margin over Astra only.
• SVG output. The author ran his pelican test at low, medium and high settings and rated all three outputs excellent. The high-setting image cost 17.166 cents. He notes that Gemini is the only lab he knows of whose staff have worked on SVG generation specifically.

Limits

• Benchmark cost figures. These come from a third party, Artificial Analysis, and the page reproduces them only in a chart. The author used ChatGPT to adjust the chart's layout. He says he checked that the data is correct.
• Per-token prices. The author argues these are poor comparators, because the cost that matters is how many tokens a model needs to finish a task inside an agentic harness. The page gives no token-count data of that kind.
• Missing practical details. The page does not state:
• the plugin install command
• the model ID string used in llm
• whether "low/medium/high" maps to a reasoning or thinking parameter
• any change in llm-gemini 0.35 other than Argon support

Check the plugin's release notes for these before you adopt it.