Gemini 4 Argon High is listed at $1.99 per Artificial Analysis Intelligence Index task, compared with $0.72 for GPT-6.1 Sol Max in the October 5, 2026 comparison snapshot. Sol has about 64% lower estimated task cost, while Argon's estimate is about 2.8 times Sol's. The figure is useful for comparing the evaluated workloads; it is not a flat API fee.
Source: @StatsWire, using Artificial Analysis’s cost per Intelligence Index task. On the chart, Argon (high) is $1.99 and marked not publicly available. GPT-6.1 Sol (max) is $0.72. Astra (max) is $3.26 and Claude Opus 5.5 (max) is $5.98. The Argon and Sol figures match the comparison table below. The chart is a wider leaderboard, not a two-model invoice.
Argon High vs Sol Max: the October 5 snapshot
The Artificial Analysis comparison uses these configurations and reports the following rounded index values and estimated task costs:
| Measure | Gemini 4 Argon High | GPT-6.1 Sol Max |
|---|---|---|
| Snapshot date | October 5, 2026 | October 5, 2026 |
| Artificial Analysis Intelligence Index | 53 | 52 |
| Estimated cost per task | $1.99 | $0.72 |
Argon is one index point higher in this snapshot, while Sol's estimated cost per task is about 36% of Argon's. The underlying Argon index record is 52.5606 before rounding to 53; the comparison displays the rounded value.
What “cost per task” means here
Artificial Analysis computes a weighted evaluation cost from the tokens and model usage needed across its Intelligence Index tasks. It is an evaluation-cost estimate, not a flat fee charged for every request, a subscription price, or a guarantee that a customer's Sol task will cost $0.72.
The ratio is also workload-dependent. A task that needs more context, longer output, retries, or tool calls can have a different cost than the weighted mix behind the index. Use the snapshot to ask whether Argon's quality difference is worth testing against your own workload.
Token rates are a separate comparison
The same Artificial Analysis comparison lists matching token rates for these configurations:
| Price item | Gemini 4 Argon High | GPT-6.1 Sol Max |
|---|---|---|
| Input / output per 1M tokens | $2 / $10 | $2 / $10 |
| Cached input per 1M tokens | $0.10 | $0.10 |
Matching token rates do not imply matching cost per completed task. Different models can use different numbers of tokens and retries to finish the same evaluation.
Which model should you test first?
Use Sol as the first candidate when the priority is lower estimated evaluation cost and a one-point index difference is acceptable for your workload. Test Argon when the extra task cost may buy enough quality or reliability to reduce retries and review time. That is a practical inference from the snapshot, not a guarantee about every project.
Argon's announced introductory API rates are also $2 per million input tokens and $10 per million output tokens, according to Google's announcement. The Argon release page explains its staged access; an Artificial Analysis task estimate does not mean the model is available to every caller.
Related reading
Try Gemma 4 online
~/gemma4 $ Try Gemma 4 in the browser playground before setting up a local installation.
Open Gemma 4 playground />
