Claude Opus 5: what changes vs Opus 4.8 and when to use it (2026)

July 25, 20269 minClaude Opus 5, Anthropic, AI, LLM, Coding

Short answer (60 seconds): Claude Opus 5 is Anthropic's newest Opus tier model (launched July 24, 2026). Same price as Opus 4.8 ($5/MTok input, $25/MTok output), 1M context, and thinking enabled by default. Leads SWE-bench Verified at 97.00%. Pick it when: agentic coding and knowledge work where the 97% vs 95% benchmark gap doesn't translate to business value — you get almost what Fable 5 offers at half the price. Don't pick it when: you need self-hosting (open-weight M3 or Kimi K3 win) or the absolute capability ceiling regardless of cost (that's still Fable 5).

Claude Opus 5 is Anthropic's most recent Opus tier release, just two months after Opus 4.8 (May 28, 2026). This post is the guide I wish I had on announcement day: what changes, what stays the same, what it actually costs, and the practical question — for which workloads does migration make sense, and for which doesn't it?

Disclosure: I've been testing Opus 5 since Claude API access opened on July 24. It's not "daily use" yet (literally one day old), but the data below comes from real prompts I ran against Opus 5 and Opus 4.8 in parallel — not just the official blog.

What changes vs Opus 4.8

Anthropic positions Opus 5 as "close to the frontier intelligence of Claude Fable 5 at half the price." Three concrete changes back that claim up:

1. Measurably stronger agentic coding. Opus 5 leads SWE-bench Verified at 97.00%, ahead of GPT-5.6 Sol (96.20%) and Fable 5 (95.00%). Compared to Opus 4.8 (88.60%), that's an ~8.4 point jump — not incremental, but a tier change. On SWE-bench Pro (longer and messier problems), Opus 5 lands at 79.2% (Fable 5 at 80.0%, Mythos 5 leads at 80.3%). In real coding, the gap shows up most on multi-hour sessions and large repo navigation.

2. Thinking enabled by default. Through Opus 4.8, extended thinking was opt-in: you had to pass thinking={"type": "enabled", "budget_tokens": N} on each call. In Opus 5 it's on by default, with adaptive budget based on prompt complexity. That changes how the model behaves on multi-step tasks without you touching the code.

3. More stable long-running agents. Anthropic reports explicit improvements on multi-hour agentic sessions with minimal oversight. Combined with the 1M context window, you can hand it an entire codebase and maintain thread continuity across long sessions without compaction.

There's a fourth change that isn't about capability but matters: reinforced cybersecurity safeguards. Anthropic strengthened these after conversations with government regulators about dual-use concerns. In practice, you'll see more rate limiting and more refusal patterns on ambiguous prompts. Not a deal-breaker, but worth knowing if you're coming from Opus 4.8.

Specs and pricing

Pricing is identical to Opus 4.8 — Anthropic didn't raise rates to fold in the improvements. To put it in perspective, Opus 5 costs half what Fable 5 costs ($10/$50 MTok) and a quarter of Mythos 5 (the top tier above Fable).

Availability by plan:

• Claude Max: Opus 5 becomes the default model. If you had Opus 4.8 configured, migrate to claude-opus-5 to get default thinking and better coding.
• Claude Pro: Opus 5 is the strongest model available in the Pro tier.
• Claude Team and Enterprise: available with the governance you already have configured.
• Claude API and Amazon Bedrock: model claude-opus-5 since July 24.

How to use Opus 5 with the Anthropic SDK

The Python SDK didn't change — only the model ID does. Here's the minimum to get started:

import os
from anthropic import Anthropic

client = Anthropic(api_key=os.environ.get("ANTHROPIC_API_KEY"))

message = client.messages.create(
    model="claude-opus-5",  # used to be "claude-opus-4-8"
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Refactor this module to use async/await instead of callbacks."}
    ],
)

print(message.content[0].text)

If you were using opt-in extended thinking (with thinking={"type": "enabled", ...}), you can drop the block: Opus 5 already thinks by default. If you want explicit budget control, the parameter still works:

response = client.messages.create(
    model="claude-opus-5",
    max_tokens=16000,
    thinking={
        "type": "enabled",
        "budget_tokens": 10000,  # explicit override of the adaptive default
    },
    messages=[{"role": "user", "content": "Solve this scheduling problem..."}],
)

Streaming works the same as before, and the thinking block shows up in response.content with type="thinking" before the final text block. If you had logic reading thinking blocks, nothing changes.

For more SDK and endpoint details, the official Anthropic docs are up to date.

Benchmarks: where Opus 5 is frontier and where it isn't yet

Public data as of July 25, 2026, comparing the strongest models available:

Coding (SWE-bench Verified):

Opus 5 leads coding by a clear margin, but the gap to Fable 5 (1.8 points) and GPT-5.6 Sol (0.8 points) is small. In real agentic coding, that gap rarely translates into additional productivity — all three are in the "good enough that you don't worry about the model" cluster.

SWE-bench Pro (longer and messier): Mythos 5 leads at 80.3%, Fable 5 at 80.0%, Opus 5 at 79.2%. Here Opus 5 is already 0.8 points behind Fable 5 — the gap starts showing on multi-hour tasks where problems accumulate.

Frontier-bench v0.1 (general reasoning): Opus 5 reaches 43.3% pass rate. I don't have exact Fable 5 and Mythos 5 numbers on this benchmark at the close of this post, but the trend is the same — Opus 5 is close to the ceiling but doesn't reach it.

What Opus 5 still doesn't reach vs Fable 5:

• Pure reasoning on very long problems (Fable 5 maintains coherence better on 4+ hour sessions)
• Professional knowledge work where the gap between "97% correct" and "99% correct" is material (legal, medical research, fine financial analysis)
• Inference speed under massive load (Fable 5 has more optimized infra)

On agentic coding and standard knowledge work, Opus 5 is indistinguishable from Fable 5 for 95% of workloads. The gap shows up only in the last 5%, where Fable 5's extra cost starts to make sense.

When to pick Opus 5 — and when not

Pick Opus 5 if:

• Your workload is agentic coding, refactoring, or standard knowledge work
• You want to keep Opus 4.8's cost ($5/$25) but step up a tier of capability
• You're on Claude Max and want the updated default model
• You need 1M context without compaction to pass entire codebases
• Anthropic's ecosystem (SDK, integrations, governance) works for you

Pick Claude Fable 5 if:

• Your workload is pure multi-step reasoning on very long sessions (4+ hours)
• Professional knowledge work where the last 1-2% of accuracy is material
• Cost isn't a concern and you want Anthropic's absolute ceiling
• You need aggressive prompt caching (Fable 5 has cache hits at $1/MTok)

Pick MiniMax M3 if:

• You need self-hosting for compliance, cost, or data control
• Your product is inherently multimodal (video, audio + text)
• You want to avoid vendor lock-in to Anthropic
• Token volume justifies operating your own infra (dedicated H100)

Pick Kimi K3 if:

• You want open weights with more long-horizon coding maturity than M3
• Your workload is primarily text (you don't need M3's native multimodality)
• You can wait until July 27 for open weights without paying API

Pick GPT-5.6 Sol if:

• Your stack is already on OpenAI and switching creates friction
• You need OpenAI-ecosystem-specific features (Assistants API, GPT Store, etc.)

The LATAM angle: what changes with Opus 5

For developers and companies in LATAM, Opus 5 matters for three concrete reasons:

1. Better price/performance ratio on managed API. LATAM still pays USD per token like the rest of the world, but the gap between $5/MTok (Opus 5) and $10/MTok (Fable 5) shows up in bills at medium volume. For teams of 5-20 engineers running coding agents, migrating from Opus 4.8 to Opus 5 gets them ~8 extra SWE-bench points at no cost change.

2. Claude Max becomes the default model. If many Spanish-speaking developers are on Claude Max (likely, given the relatively accessible pricing vs going direct on Fable 5), Opus 5 reaches them automatically. Nothing to do — just verify the updated model ID is in use.

3. Default thinking reduces friction. In LATAM, where many teams adopt AI later than in the US, opt-in thinking was a barrier: you had to know it existed, read the docs, configure it. With Opus 5 it's on by default, which lowers the learning curve for teams just starting with agents.

What Opus 5 doesn't solve for LATAM:

• Data compliance. Opus 5 is still a managed API on Anthropic's servers. If you need data in-region (local regulations), self-hosting with M3 or Kimi K3 is still the only path.
• Cost at high volume. Self-hosting with open weights is still 30-50% cheaper than any managed API past ~50M tokens/month.

Want to discuss whether Opus 5 makes sense for your stack before migrating? There's a CTA at the end with a free 30-minute call.

Frequently asked questions

What is Claude Opus 5 in one sentence?

It's Anthropic's newest Opus tier model, launched July 24, 2026. Same price as Opus 4.8 ($5/MTok input, $25/MTok output), 1M context, thinking enabled by default, and it lands close to Claude Fable 5's performance at half the cost.

What changes vs Opus 4.8?

Three main changes: (1) measurably stronger agentic coding — leads SWE-bench Verified at 97.00% vs Opus 4.8's 88.60%; (2) thinking enabled by default instead of opt-in, which changes how it behaves on multi-step tasks; (3) improvements in long-running agents and cybersecurity safeguards per regulator requests. Context window and pricing stay the same.

How much does it cost?

$5 per million input tokens and $25 per million output tokens — identical to Opus 4.8. Available on Claude API, Amazon Bedrock, and the Pro, Max, Team, and Enterprise Claude plans. On Max it becomes the default model; on Pro it's the strongest available.

When does Opus 5 beat Claude Fable 5?

Fable 5 stays Anthropic's top tier (launched June 9, 2026 at $10/$50 MTok) and leads on pure reasoning and professional agentic work. If your workload is agentic coding or knowledge work where the 97% vs 95% SWE-bench gap doesn't translate to business value, Opus 5 gets you almost the same at half the price. Pick Fable 5 only if you need the absolute capability ceiling and cost is not a concern.

When does Opus 5 beat MiniMax M3 or Kimi K3 (open-weight)?

If you want a managed API without operating infra, or you need Anthropic's polished UX and integration ecosystem. If you need self-hosting for LATAM data compliance, high-volume cost control, or vendor lock-in avoidance, open-weight wins: M3 with native multimodality, Kimi K3 with its open-weight release on July 27.

Does Opus 5 work well in Spanish?

Yes. Anthropic's multilingual training handles Latin American Spanish prompts well, same as Opus 4.8 and Fable 5. For domain-specific cases (LATAM legal, contracts with local terminology) I'd recommend testing with your own prompts before committing to a large workload.