# Codex vs Claude Code, measured side by side (plus Antigravity CLI)

As of 2026-11-11 14:38 UTC. The same fixed task every hour for every agent, from one EU server, on the plans we have: Claude Code on the Claude Max subscription, Codex CLI on the ChatGPT Free plan and Antigravity CLI on the Google AI subscription. Only figures measured the same way for all.

As of 2026-11-11 14:38 UTC, median time to answer the same fixed task, end to end, over the last 24 hours — Claude Code with Claude Opus 5.5: 15.4 s (24 runs); Claude Code with Claude Sonnet 5.5: 9.2 s (3 runs, too few to compare); Codex CLI with GPT-6 Luna: 25.7 s (4 runs, too few to compare); Antigravity CLI with Gemini 3.8 Flash: 99.3 s (24 runs); Antigravity CLI with GPT-OSS 120B: 16.4 s (3 runs, too few to compare). Claude Code vs Codex CLI over the last 7 days: too few runs to call it yet (167 successful runs of Claude Code with Claude Opus 5.5 and 4 of Codex CLI with GPT-6 Luna in 7 days). Failed calls in the last 7 days — Claude Code with Claude Opus 5.5: 0 of 167; Claude Code with Claude Sonnet 5.5: 0 of 3; Codex CLI with GPT-6 Luna: 0 of 4; Antigravity CLI with Gemini 3.8 Flash: 0 of 167; Antigravity CLI with GPT-OSS 120B: 0 of 3. On the 20 hard questions both answered in the last 7 days, Claude Code with Claude Opus 5.5 got 64% right and Codex CLI with GPT-6 Luna 45% — no significant difference (+18.6 points ± 11.9, z = 1.57).

## Time to answer the same task

| Agent · model | Plan | Median, last 24 h | Median, last 7 days | Failed, 24 h | Failed, 7 days |
|---|---|---|---|---|---|
| Claude Code · Claude Opus 5.5 | Claude Max subscription | 15.4 s (24 runs) | 15.7 s (167 runs) | 0 of 24 | 0 of 167 |
| Claude Code · Claude Sonnet 5.5 | Claude Max subscription | 9.2 s (3 runs, too few to compare) | 9.2 s (3 runs, too few to compare) | 0 of 3 | 0 of 3 |
| Codex CLI · GPT-6 Luna | ChatGPT Free plan | 25.7 s (4 runs, too few to compare) | 25.7 s (4 runs, too few to compare) | 0 of 4 | 0 of 4 |
| Antigravity CLI · Gemini 3.8 Flash | Google AI subscription | 99.3 s (24 runs) | 102.6 s (167 runs) | 0 of 24 | 0 of 167 |
| Antigravity CLI · GPT-OSS 120B | Google AI subscription | 16.4 s (3 runs, too few to compare) | 16.4 s (3 runs, too few to compare) | 0 of 3 | 0 of 3 |

## Pairs, last 7 days

- Claude Code with Claude Opus 5.5 vs Codex CLI with GPT-6 Luna: time — too few runs to call it yet (167 successful runs of Claude Code with Claude Opus 5.5 and 4 of Codex CLI with GPT-6 Luna in 7 days); accuracy — on the 20 hard questions both answered in the last 7 days, Claude Code with Claude Opus 5.5 got 64% right and Codex CLI with GPT-6 Luna 45% — no significant difference (+18.6 points ± 11.9, z = 1.57).
- Claude Code with Claude Opus 5.5 vs Claude Code with Claude Sonnet 5.5: time — too few runs to call it yet (167 successful runs of Claude Code with Claude Opus 5.5 and 3 of Claude Code with Claude Sonnet 5.5 in 7 days); accuracy — on the 20 hard questions both answered in the last 7 days, Claude Code with Claude Opus 5.5 got 64% right and Claude Code with Claude Sonnet 5.5 55% — no significant difference (+8.6 points ± 12.1, z = 0.71).
- Claude Code with Claude Opus 5.5 vs Antigravity CLI with Gemini 3.8 Flash: time — Claude Code with Claude Opus 5.5 was faster: Antigravity CLI with Gemini 3.8 Flash took 6.5× as long (15.7 s vs 102.6 s median, p < 0.001); accuracy — on the 76 hard questions both answered in the last 7 days, Claude Code with Claude Opus 5.5 got 60% right and Antigravity CLI with Gemini 3.8 Flash 55% — no significant difference (+5.4 points ± 5.1, z = 1.07).
- Claude Code with Claude Opus 5.5 vs Antigravity CLI with GPT-OSS 120B: time — too few runs to call it yet (167 successful runs of Claude Code with Claude Opus 5.5 and 3 of Antigravity CLI with GPT-OSS 120B in 7 days); accuracy — on the 38 hard questions both answered in the last 7 days, Claude Code with Claude Opus 5.5 got 69% right and Antigravity CLI with GPT-OSS 120B 45% — Claude Code with Claude Opus 5.5 was more accurate (+24.0 points ± 7.6, z = 3.16).
- Claude Code with Claude Sonnet 5.5 vs Codex CLI with GPT-6 Luna: time — too few runs to call it yet (3 successful runs of Claude Code with Claude Sonnet 5.5 and 4 of Codex CLI with GPT-6 Luna in 7 days); accuracy — on the 20 hard questions both answered in the last 7 days, Claude Code with Claude Sonnet 5.5 got 55% right and Codex CLI with GPT-6 Luna 45% — no significant difference (+10.0 points ± 10.0, z = 1.00).
- Claude Code with Claude Sonnet 5.5 vs Antigravity CLI with Gemini 3.8 Flash: time — too few runs to call it yet (3 successful runs of Claude Code with Claude Sonnet 5.5 and 167 of Antigravity CLI with Gemini 3.8 Flash in 7 days); accuracy — on the 20 hard questions both answered in the last 7 days, Claude Code with Claude Sonnet 5.5 got 55% right and Antigravity CLI with Gemini 3.8 Flash 57% — no significant difference (-2.5 points ± 11.7, z = -0.21).
- Claude Code with Claude Sonnet 5.5 vs Antigravity CLI with GPT-OSS 120B: time — too few runs to call it yet (3 successful runs of Claude Code with Claude Sonnet 5.5 and 3 of Antigravity CLI with GPT-OSS 120B in 7 days); accuracy — on the 11 hard questions both answered in the last 7 days, Claude Code with Claude Sonnet 5.5 got 64% right and Antigravity CLI with GPT-OSS 120B 45% — too few common questions to compare yet.
- Codex CLI with GPT-6 Luna vs Antigravity CLI with Gemini 3.8 Flash: time — too few runs to call it yet (4 successful runs of Codex CLI with GPT-6 Luna and 167 of Antigravity CLI with Gemini 3.8 Flash in 7 days); accuracy — on the 20 hard questions both answered in the last 7 days, Codex CLI with GPT-6 Luna got 45% right and Antigravity CLI with Gemini 3.8 Flash 57% — no significant difference (-12.5 points ± 12.5, z = -1.00).
- Codex CLI with GPT-6 Luna vs Antigravity CLI with GPT-OSS 120B: time — too few runs to call it yet (4 successful runs of Codex CLI with GPT-6 Luna and 3 of Antigravity CLI with GPT-OSS 120B in 7 days); accuracy — on the 11 hard questions both answered in the last 7 days, Codex CLI with GPT-6 Luna got 36% right and Antigravity CLI with GPT-OSS 120B 45% — too few common questions to compare yet.
- Antigravity CLI with Gemini 3.8 Flash vs Antigravity CLI with GPT-OSS 120B: time — too few runs to call it yet (167 successful runs of Antigravity CLI with Gemini 3.8 Flash and 3 of Antigravity CLI with GPT-OSS 120B in 7 days); accuracy — on the 37 hard questions both answered in the last 7 days, Antigravity CLI with Gemini 3.8 Flash got 57% right and Antigravity CLI with GPT-OSS 120B 46% — no significant difference (+10.8 points ± 7.3, z = 1.48).

Rule: only end-to-end time on the same fixed task (from launching the CLI to the complete answer, median of successful calls; comparable once it has >= 12 calls in 24 h or >= 48 in 7 days; two agents differ when their 7-day medians are >= 15% apart and a Mann-Whitney test gives p < 0.01), the share of failed calls, and hard-panel accuracy of each pair on the questions both answered in the last 7 days (paired per-question difference, item-clustered SE; significant when |delta| >= 5 points and |z| >= 2.58 with >= 20 common questions). Tokens/s, thinking and time to first token are not compared across agents: Claude Code reports API time, Codex and Antigravity CLI turn time, and the tokenizers differ.

## Questions

### Is Codex faster than Claude Code?

On the same fixed task over the last 7 days: too few runs to call it yet (167 successful runs of Claude Code with Claude Opus 5.5 and 4 of Codex CLI with GPT-6 Luna in 7 days). Over the last 24 hours Claude Code with Claude Opus 5.5 took a median 15.4 s (24 runs) and Codex CLI with GPT-6 Luna 25.7 s (4 runs, too few to compare). Claude Code runs on the Claude Max subscription, Codex on the ChatGPT Free plan.

### Is Claude Code or Codex more accurate?

On the 20 hard questions both answered in the last 7 days, Claude Code with Claude Opus 5.5 got 64% right and Codex CLI with GPT-6 Luna 45% — no significant difference (+18.6 points ± 11.9, z = 1.57). The 78-question panel was selected at the edge of Claude Opus 5.5's ability (questions it solves only sometimes), so it is not a neutral ranking of models.

### Which coding agent fails more often?

Failed calls on the same hourly task in the last 7 days — Claude Code with Claude Opus 5.5: 0 of 167; Claude Code with Claude Sonnet 5.5: 0 of 3; Codex CLI with GPT-6 Luna: 0 of 4; Antigravity CLI with Gemini 3.8 Flash: 0 of 167; Antigravity CLI with GPT-OSS 120B: 0 of 3. A call counts as failed when the CLI errors out, times out or is rate-limited, or cannot count as a clean measurement.

### Why not compare tokens per second or thinking tokens?

Tokens per second, thinking tokens and time to first token are not compared between agents: Claude Code reports API time while Codex and Antigravity CLI report turn time, Antigravity CLI does not stream thinking, and the tokenizers differ, so the same number means different things. Each agent's own page judges them against that agent's own last 7 days.

### Which models and plans are compared?

Claude Code (Claude Opus 5.5), Codex CLI (GPT-6 Luna), Antigravity CLI (Gemini 3.8 Flash), Claude Code (Claude Sonnet 5.5) and Antigravity CLI (GPT-OSS 120B); Claude Code on the Claude Max subscription, Codex CLI on the ChatGPT Free plan and Antigravity CLI on the Google AI subscription. Each runs on one account from one EU server; plan limits, routing and default models are part of what is measured.

Per-agent pages: [Is Claude Code down or slower today?](https://llmlatency.dev/coding-agents/claude-code) · [Is Codex down or slower today?](https://llmlatency.dev/coding-agents/codex) · [Is Antigravity CLI down or slower today?](https://llmlatency.dev/coding-agents/antigravity) · [all coding agents](https://llmlatency.dev/coding-agents).

Data: [/api/coding-agents.json](https://llmlatency.dev/api/coding-agents.json) (JSON, CC BY 4.0).