Claude Opus 4.8 Speed Test — First Look at Performance Data
Anthropic just shipped Claude Opus 4.8. We added it to ModelStats on day one. Here's the early TTFT, throughput, and latency data versus Opus 4.7, GPT-4o, and Gemini.
Anthropic released Claude Opus 4.8 on May 28, 2026 — less than two months after Opus 4.7. We added it to ModelStats the same day and started pinging it every 10 minutes.
Here's what we're seeing.
What's New
Claude Opus 4.8 is Anthropic's new flagship. Like the 4.7 release before it, this is an Opus-only update — Sonnet 4.6 and Haiku 4.5 are unchanged. Anthropic positions 4.8 as having "sharper judgement, more honesty about its progress, and the ability to work independently for longer than its predecessors."
The headline numbers from Anthropic:
- Agentic coding climbs from 64.3% to 69.2%
- Fast mode is roughly 2.5× quicker than before
- Pricing is unchanged from Opus 4.7
Those are quality and capability claims. ModelStats only measures one thing: real-world API speed. So how fast is it actually responding?
What We're Measuring
Every 10 minutes, we send Opus 4.8 a streaming request and measure:
- TTFT (Time to First Token) — how long until the model starts responding
- Total Latency — full round-trip time
- Throughput (Tokens/s) — how fast it generates output
- ITL (Inter-Token Latency) — streaming smoothness
- Error Rate — reliability
Early Performance Data
Check the live dashboard for the most current numbers — these are from the first hours of monitoring.
In early testing, Opus 4.8 is trending slightly ahead of 4.7 on TTFT and throughput, which lines up with Anthropic's claim that the fast-mode path got a big speed boost. As always, the picture sharpens over the coming days as we collect more samples across different times of day and load conditions.
How It Compares
The Anthropic lineup now looks like:
| Model | Role | Status |
|-------|------|--------|
| Claude Opus 4.8 | Flagship, most capable | New |
| Claude Opus 4.7 | Previous flagship | Still available |
| Claude Sonnet 4.6 | Balanced performance | Unchanged |
| Claude Haiku 4.5 | Fastest, cheapest | Unchanged |
Opus is the quality leader in the lineup, not the speed leader — if raw latency is your priority, Haiku 4.5 and Sonnet 4.6 remain the faster Anthropic options. The interesting question is whether 4.8's fast-mode improvements narrow that gap.
Cross-Provider Comparison
See how Opus 4.8 stacks up against the competition:
- vs GPT-4o: Check the head-to-head comparison
- vs Gemini 2.5 Flash: Check the comparison
- vs Opus 4.7: See both on the live dashboard to compare generations side by side
Should You Upgrade?
If you're on Opus 4.7, the upgrade is a one-line model ID change (claude-opus-4-7 → claude-opus-4-8) and pricing is identical, so there's little downside to testing it. Monitor the performance data on ModelStats for a few days to confirm the speed profile holds for your traffic patterns before flipping production over.
If you're choosing between providers, use our scatter plot comparison, which plots TTFT vs throughput across all 15 models we track.
Key Takeaways
- Claude Opus 4.8 is available now — Opus only, not Sonnet or Haiku
- Same pricing as 4.7, with stronger agentic coding and a much faster fast mode
- Early ModelStats data shows TTFT and throughput trending ahead of 4.7
- We're tracking it alongside 14 other models with real-time data
All data is from real API monitoring at modelstats.ai, updated every 10 minutes.