Claude Sonnet 5: Release Tracker

OK so I've been glued to my Vertex AI dashboards since last weekend, and honestly? This one caught me off guard. A claude-sonnet-5@20260203 entry surfaced in what looks like a misconfigured Google Cloud deployment log, and the developer community hasn't stopped talking about it since. I've been tracking Anthropic's release patterns for over two years now — I know the difference between a test artifact and something real — and the signals here are unusually strong. So I built this page to be your single source of truth. Everything below is either pulled directly from Anthropic's official docs or sourced from verified hands-on reports. I'll keep updating it as things move.

Official Sources to Watch

Before chasing rumors, bookmark these. These are the only places where a Claude Sonnet 5 release will be officially confirmed — and where model IDs, pricing, and capabilities will first appear.

Here's a quick way to programmatically check what's actually live on your end right now — no guessing needed:

import anthropic

client = anthropic.Anthropic()  # uses ANTHROPIC_API_KEY from env

models = client.models.list()
for model in models.data:
    print(f"{model.id:<45} | released: {model.created_at}")

Run this against your own API key. If claude-sonnet-5-20260203 shows up in the output, it's live on Anthropic's first-party API for your account. I checked this morning — not there yet on production. But that's exactly the kind of thing that changes fast.

For AWS Bedrock users, the equivalent CLI check:

aws bedrock list-foundation-models \
  --region us-west-2 \
  --by-provider anthropic \
  --query "modelSummaries[*].modelId"

Rollout Status by Provider / Region

This is where things get practical. Anthropic doesn't flip a global switch — models roll out across providers and regions in stages. I've mapped the current state based on the official endpoint documentation and live testing.

Current Confirmed Lineup (as of Feb 3, 2026)

All IDs above (except Sonnet 5) are verified directly from the Anthropic models overview page.

US vs EU vs APAC Availability Notes

This is the part most guides skip, and it matters if you're building in production. Starting with Claude Sonnet 4.5, Anthropic introduced a two-tier endpoint structure on both Bedrock and Vertex AI:

• Global endpoints — dynamic routing for maximum availability. This is the default and the recommended path for most teams. On Bedrock, the model ID includes a global. prefix: global.anthropic.claude-sonnet-4-5-20250929-v1:0. On Vertex AI, set region = "global".
• Regional endpoints — guaranteed data routing within a specific geographic boundary. These carry a 10% pricing premium over global endpoints (confirmed in Anthropic's pricing docs). Use these only when you have strict data residency requirements.

Here's what that looks like in code on Bedrock — the difference is literally one prefix:

from anthropic import AnthropicBedrock

client = AnthropicBedrock(aws_region="us-west-2")

# Global endpoint (default, recommended)
msg_global = client.messages.create(
    model="global.anthropic.claude-sonnet-4-5-20250929-v1:0",
    max_tokens=256,
    messages=[{"role": "user", "content": "Hello"}]
)

# Regional endpoint (US only, strict data residency)
msg_regional = client.messages.create(
    model="anthropic.claude-sonnet-4-5-20250929-v1:0",  # no global. prefix
    max_tokens=256,
    messages=[{"role": "user", "content": "Hello"}]
)

For EU and APAC teams: regional availability varies. The safest approach right now is to hit the global endpoint and let Anthropic's routing layer handle it. If you're in Tokyo or Frankfurt and seeing latency spikes, that's when you switch to a pinned regional endpoint and test from there.

When Sonnet 5 drops, expect the same pattern: global first, regional rollout within days. I'll update the table above the moment it's confirmed.

Change Log

This section is timestamped. Newest entries first. Everything here is sourced — I don't add speculation without clearly labeling it.

What to actually do right now

I know the instinct is to wait. But here's my honest take from running production workloads on Claude for a while: the best time to prepare for a new model is before it drops, not after.

First, make sure your model selection isn't hardcoded. Use Anthropic's aliased model names (claude-sonnet-4-5 instead of claude-sonnet-4-5-20250929) for anything that should auto-upgrade — but keep pinned snapshot versions in production until you've tested the new model yourself. Second, if Sonnet 5 really does land at the same $3/$15 price point as Sonnet 4.5, the migration cost is basically zero. The risk of not evaluating it quickly is much higher than the risk of a test run.

I'll keep this page updated. If you're building on Claude and want to be notified the second something changes, follow the Anthropic Release Notes — that's ground zero.