OpenCode outperforms Gemini CLI using Gemini models. We benchmark them head-to-head on Gemini 3 Flash Preview and Gemini 3 Pro Preview to see how a vendor-agnostic agent compares to Google’s own Gemini CLI.

What we measured:

• Sigmascore — the overall measure of an agent’s real-world coding performance
• Accuracy — how often outputs meet quality thresholds
• Consistency — how often outputs remain useful even when not fully completing a task
• Speed — how quickly tasks are completed

Each score is assigned a tier based on how close they are with respect to the margin of error. The top-scoring group of agents are in Tier 1, the next-best group are in Tier 2, and so on.

Across Gemini 3 Flash Preview and Gemini 3 Pro Preview, OpenCode outperforms Gemini CLI:

• It scores higher on every metric of our benchmark across both runs.
• It’s faster in both configurations (31% faster on Flash, 24% faster on Pro).
• OpenCode (Gemini 3 Flash Preview) ranks #2 overall on our leaderboard.

Comparisons

OpenCode vs Gemini CLI (Gemini 3 Flash Preview)

OpenCode comes out on top here. It beats Gemini CLI on every metric of our benchmark. Strikingly, OpenCode sits a full tier higher on our composite Sigmascore metric, highlighting a categorical performance advantage.

Accuracy and Consistency are both strong, and with a median runtime 31% lower than Gemini CLI.

OpenCode vs Gemini CLI (Gemini 3 Pro Preview)

When using Gemini 3 Pro Preview, OpenCode comes out ahead again. Even more astonishingly, OpenCode sits two full tiers higher on our composite Sigmascore metric compared to Gemini CLI.

OpenCode is also 24% faster, with a median runtime of 325s vs 426s.

Median Runtime Comparison

The following chart shows the dramatic runtime gap between OpenCode and Gemini CLI across both Gemini 3 models. Note that OpenCode is faster using Gemini 3 Pro Preview than Gemini CLI is using Gemini 3 Flash Preview!

Notes

Thinking Level / Reasoning Effort

In order to keep the comparison fair, we matched model settings across both agents where possible. Gemini CLI defaults to the high thinking level when using Gemini 3 models, so we matched it in OpenCode by running with --variant high.

Gemini CLI Early Exits (Loop Detection)

We observed cases where Gemini CLI triggered its loop detection and exited early, which we counted as failed runs (12.1% for Flash, 11% for Pro). This behavior is reported in an open issue, and may contribute to some of Gemini CLI’s underperformance in this report. We plan to rerun and update the numbers once this is resolved.

OpenCode vs Gemini CLI - Key Insights

• OpenCode scores higher on every Sigmabench metric across both Gemini 3 models.

• OpenCode is consistently faster than Gemini CLI: 31% faster on Flash, 24% faster on Pro.

• Largest gap is with Gemini 3 Pro Preview: OpenCode sits two tiers higher on Sigmascore.