Where it stands
Anthropic has announced Claude Opus 5.5 as the first model in a Claude 5.5 family, according to its official X post. The company says it performs at Claude Fable 5.1's level for most tasks and costs 40% less to run than Opus 5; those comparisons are not independently verified. Earlier reports, including a Claude Code change entry, a stealth-testing post and per-million-token prices of $4 input and $20 output, remain unverified and are not tied to official pricing. Availability, API access, pricing and benchmark details are still open.
Anthropic announces Claude Opus 5.5 as the first model in a 5.5 family
Claude Opus 5.5 is presented as the first model in a new Claude 5.5 family. In a post on X, @claudeai says it performs at the level of Claude Fable 5.1 for most tasks and costs 40% less to run than Opus 5.
Those are Anthropic's own figures, not independent results. The linked article could not be read, and the source reading is truncated, so additional details may exist outside the material available here.
Where Claude Opus 5.5 fits
Anthropic describes Claude Opus 5.5 as the first model in the Claude 5.5 family. It does not explain what other models the family will include or how Opus 5.5 differs technically from earlier Claude models.
The excerpt also compares its performance with Claude Fable 5.1. Specifically, it says Opus 5.5 performs at that model’s level for most tasks. Anthropic doesn't name benchmarks, scores, testing methods or examples of the tasks covered by “most.”
What the cost comparison means
The post says Claude Opus 5.5 costs 40% less to run than Opus 5. In percentage terms, that means the stated running cost would be 60% of the Opus 5 figure.
For example, if Opus 5 had a hypothetical running cost of 100 units, a 40% reduction would put the corresponding Opus 5.5 cost at 60 units. This illustrates the percentage only; Anthropic didn't give an absolute price or cost unit.
Still open at launch
The announcement didn't say when Claude Opus 5.5 would be available, where it could be used, whether it had API access, or what its specifications are. It also does not show benchmark evidence or a methodology for either the performance or cost comparison. Because the linked article could not be read, these are limits of the evidence available for this explanation rather than definitive statements about the full announcement.
The clearest takeaway is that Opus 5.5 is being introduced as the first Claude 5.5 model, with reported performance comparable to Claude Fable 5.1 for most tasks and a reported running cost 40% below Opus 5.
Sources
Claude Code 2.1.280 change summary lists claude-opus-5-5 and Responsive mode
A post on X from @synthwavedd says that claude-opus-5-5 has appeared in the imminent Claude Code v2.1.280. The attached screenshot appears to show a RAGtrack summary of changes between Claude Code versions 2.1.278 and 2.1.280, including a reported Opus 5.5 model entry and a new Responsive mode.
This is a report of an apparent code change, not confirmation that Claude Opus 5.5 has launched or is available to Claude Code users. Anthropic had not announced or documented the change at that point.
What the screenshot lists for Claude Opus 5.5
The screenshot labels claude-opus-5-5 as an Opus-family model and lists a 1M-token context figure alongside a 128K maximum output figure. It also shows the pricing identifier tier_4_20_cache_read_0_20, a medium default-effort setting, the fallback entry fallback_3p: claude-opus-5, the Bedrock model identifier us.anthropic.claude-opus-5-5, and the Vertex environment variable VERTEX_REGION_CLAUDE_5_5_OPUS.
Those entries appear in the change summary, but their presence does not establish actual user pricing, practical availability of a 1M-token context window, or model performance. The screenshot also lists capability flags and descriptive picker text, including a “best for complex tasks” option and a description referring to long sessions with large codebases. These are presented as configuration or interface details rather than independent evidence of the model’s capabilities.
The summary appears to position Opus 5.5 as the new top Opus model and to describe Opus 5 as a previous version. That wording indicates how the reported change is framed, but it does not confirm a completed public release.
What the reported Responsive mode does
The screenshot describes Responsive mode as a mode in which Claude replies with a sentence before thinking or using tools on every prompt. It also lists the internal-looking tengu_quiet_ember value and a built-in output style described as clear, conversational English.
It wasn't clear how users would enable Responsive mode, whether it was available beyond the reported change, or whether the behavior applied in every Claude Code setup. The mode should therefore be understood as a reported change-summary entry, not as documented usage guidance.
Other changes shown for Claude Code
The same summary lists several changes alongside the model entry and Responsive mode:
Browser handling: New browser-selection items, including
get_chrome_browsers,select_chrome_browser, and paired-browser naming. It also lists a hiddenclaude edit-chrome-settingssubcommand that reads a JSON edit from standard input for the VS Code extension.Sandbox settings: A new
claude edit-sandbox-settingssubcommand is listed.Claude sessions: The summary describes a device re-registration flow for a computer removed from a user’s devices. It also lists identifiers named
target_device_unsupportedandinvalid_target_device_id, plustengu_device_reenroll.Environment variables: The visible portion includes
CLAUDE_CODE_PLUGIN_DIRS,CLAUDE_CODE_INLINE_TOOLS,CLAUDE_CODE_HOST_SCHEDULED_RUN, andCLAUDE_CODE_MAX_MCP_DESCRIPTION_LENGTH.
The attached image ends while the environment-variable section is still visible, so it may not show the full list of changes for the reported version transition.
What remains unconfirmed
At that point there was no confirmation that Claude Code v2.1.280 had been released, that claude-opus-5-5 could be selected by users, or that Anthropic had announced Claude Opus 5.5. It also does not verify the listed context and output figures, explain the pricing identifier in terms of user billing, or provide benchmark results.
For now, the most supportable description is that an apparent Claude Code change summary includes a claude-opus-5-5 entry and a Responsive mode. Whether those entries pointed to an imminent release, an internal configuration or a feature users could access was still unclear.
Sources
Stealth-testing report names the claude-wafer-eap codename and a Tuesday release
A post on X by @lyraxana reports that Anthropic is stealth testing a model identified as Claude Opus 5.5 under the reported codename claude-wafer-eap. The post also says a release is planned for Tuesday.
It does not establish that Anthropic has officially announced Claude Opus 5.5, that the model is publicly available, or that a release will happen on that schedule.
What the post reports about Claude Opus 5.5
The post identifies the reported model as claude-opus-5-5 and describes Anthropic’s testing as “stealth” testing. In this context, that wording suggests testing Anthropic had not publicly documented. It does not explain how the model was accessed, how long testing has been underway, or whether the model is being evaluated internally or with a limited group of users.
The post does not provide capabilities, benchmark results, pricing, access terms, supported products, or a comparison with an existing Claude model.
The reported testing codename
The reported testing identifier is claude-wafer-eap. The post associates that name with the reported testing of claude-opus-5-5, but it does not explain the label’s meaning.
It is therefore not clear whether claude-wafer-eap is a testing label, an early-access identifier, or a name that could appear in a public product. There was no sign yet that it would become the final name of any Anthropic model or feature.
What “Tuesday” does—and does not—tell us
The post says the release is planned for Tuesday, but it does not specify which Tuesday or provide an exact calendar date. “Planned” describes an intended schedule rather than a completed release, so the statement should not be treated as confirmation that the model launched or will be generally available on that day.
The source also does not say what a release would involve. It could refer to a public model launch, access for a limited group, an API update, or another form of availability, and the post doesn't say which.
What remains unconfirmed
At that point there was no confirmation of:
An official Anthropic announcement of Claude Opus 5.5
A specific release date
Public availability or limited-access availability
The model’s capabilities or benchmark performance
Pricing, usage limits, or access requirements
Which Claude products or APIs might support it
Whether
claude-wafer-eapis a testing identifier or a public product name
Sources
Reported Opus 5.5 token prices: $4 input and $20 output per million
A post from @synaptized reports four Opus 5.5 pricing figures: $4 per million input tokens, $20 per million output tokens, $0.20 for cache reads and $5 for cache writes. The post explicitly gives the input and output figures on a per-million-token basis, but does not specify the billing unit for either cache figure. These figures come from a single X post, not an official pricing page, so treat them as reported rather than confirmed pricing.
Opus 5.5 pricing reported in the post
The post lists these figures:
Input: $4 per million tokens
Output: $20 per million tokens
Cache read: $0.20, with the billing unit unspecified
Cache write: $5, with the billing unit unspecified
The source labels cache read and cache write as separate categories but does not explain how they are measured or charged. In particular, it isn't clear that the cache figures use the same per-million-token unit as the input and output figures.
What the listed figures show
Because the post gives both input and output prices per million tokens, the reported output rate is five times the reported input rate. That comparison applies only to those two explicitly labelled per-million-token figures.
The cache-read and cache-write amounts can't be compared directly with the input and output rates because their billing units aren't stated. The figures also do not establish the cost of a particular request: the post provides no token-usage example or broader billing rules.
Not yet confirmed
The post does not identify an official provider or link to an official pricing page. It also does not establish whether Opus 5.5 is available through an API, when the figures would take effect, or what billing conditions apply.
Nothing independently confirms that the listed figures are current or apply everywhere.
Sources
Internal code reportedly cites Opus 5.5 without 5.1 or 5.2 references
A post says Anthropic’s purported internal code refers to Opus 5.5 without visible references to Opus 5.1 or Opus 5.2. That is an unverified versioning observation, not confirmation that Anthropic has announced, released or scheduled a model called Opus 5.5.
In the post, @White1637402 writes: “No 5.1, no 5.2. Anthropic's internal code just jumped straight to Opus 5.5🤔” The post doesn't show the internal code or explain how the reference was found.
What the post reports
The observation concerns a purported internal reference to the name “Opus 5.5.” It also suggests that the observer did not see references to Opus 5.1 or Opus 5.2.
It does not show that Opus 5.5 is a public model, that Anthropic has completed development, or that a release is planned. It also does not establish whether the apparent version jump reflects a real product roadmap, an internal naming convention, a test version, a temporary reference or something else.
Is Anthropic skipping Opus 5.1 and 5.2?
The post doesn't answer that. Not seeing a reference to a version in the material described by the post is not the same as proving that Anthropic has cancelled, skipped or renamed a release.
Nothing explains why the numbers might jump from 5.0 to 5.5. It also does not establish whether Opus 5.1 or Opus 5.2 were ever planned or publicly discussed. As a result, the version-number observation should not be treated as a confirmed roadmap decision.
Has Anthropic announced Opus 5.5?
Anthropic had not announced anything at that point. The post provides no release date, capabilities, benchmarks, pricing, availability details or product documentation for Opus 5.5.
Readers would need an official announcement or other reliable, corroborated evidence before treating the name as a confirmed model release. Until then, the report is best understood as a brief rumor about purported internal code rather than news of a public Claude Opus model.
0 comments
No approved comments yet. You can start the conversation.
Leave a comment