Opus 5 vs Sonnet 5 for coding
Both are current-generation Claude models with a 1M context window. Opus 5 costs 2.5× more per token. Here is what you actually get for it, and when you do not need it.
Input / M (llmrelay): Anthropic list is $5 and $2 respectively. Sonnet 5’s $2/$10 list is now Anthropic’s standard price (the 1 September 2026 increase to $3/$15 was cancelled), so our half-list price stays $1.00/$5.00.
Anthropic latency rating: Anthropic's own comparative rating, not our measurement. The scale runs Slower / Moderate / Fast / Fastest.
Reliable knowledge cutoff: Four months of library and framework churn. This is the one spec difference that bites silently — Sonnet 5 will confidently use an API that changed after January.
Prices are llmrelay's, at 50% of official list. Specs are the vendor's own published figures (Anthropic model docs), not our benchmarks. We do not publish scores we cannot source.
Use Opus 5 when
- +The task depends on recent library behaviour. The May 2026 cutoff versus January is the difference between knowing a breaking change and inventing an API that no longer exists.
- +You are running long agent trajectories where one wrong turn costs twenty tool calls. Paying 2.5× on a task that would otherwise be retried twice is cheaper than the retries.
- +You are reviewing or debugging code you did not write, where reading intent matters more than emitting volume.
Use Sonnet 5 when
- +You are doing volume: batch refactors, test generation, docstrings, migrations across many files. Anthropic rates it Fast against Opus 5's Moderate, and it costs 40% of the price.
- +The work is well specified. When you already know the answer shape, deeper reasoning buys nothing.
- +You are inside an interactive editor loop where waiting is the real cost.
The honest answer
Default to Sonnet 5 and escalate. Most people who "need Opus" need it for maybe one task in five, and the honest way to find out is to run the same prompt through both — one llmrelay key covers all of them, so switching is a one-line model-id change.
What this costs you per month
At 50M input and 5M output tokens a month — a realistic heavy agent workload.
Struck-through column is the vendor's list price for the same traffic. Adjust the numbers on the calculator.
One key covers both
Switching between them is a model-id change, not a new account. Prepaid, no subscription, $10 minimum — enough to run the same prompt through both and settle the argument with your own output.
Next steps on this decision
Per-model detail for both sides, and the adjacent comparisons people usually read alongside this one.
- Claude Opus 5 pricing and specs$2.50/M in, $12.50/M out — half list. Current flagship Opus. A generation past 4.8, same list price.
- Claude Sonnet 5 pricing and specs$1.00/M in, $5.00/M out — half list. The next-gen Sonnet. Priced to replace GPT for daily work.
- Fable 5 vs Opus 5 — is the flagship worth 2×?Fable 5 is the previous Fable snapshot, still sold at $5/$25 — twice Opus 5.
- Opus vs Sonnet for coding (measured)Measured on our own billing: $0.0152 against $0.0047 for one identical task.
- GPT-6 Astra vs Opus 5 — is 2× worth it?Astra is OpenAI's current flagship.
- V4.1 Flash vs Sonnet 5 — cheap bulk vs Anthropic defaultV4.1 Flash is the current DeepSeek Flash, GA 10 September 2026.