Claude Sonnet 5
Production workloads that need near-Opus quality with better speed and cost.
What are Claude Sonnet 5's specs and price?
Claude Sonnet 5, built by Anthropic, ships a 500K-token context window and a 64K-token max output, released 2026-06. It supports text and vision input with a dedicated reasoning mode and costs $4.00 per million blended tokens, the 32nd-cheapest of 39 models we track.
Batch 41 evidence surface · verified 2026-08-27 · exact route allowlist: /models/claude-sonnet-5
Claude Sonnet 5 production-control and agent architecture evidence
Batch 41 · M1: Identity-and-control acceptance matrix
Formula: Accepted = identity pinned ∧ requested controls accepted ∧ effective response fields present; missing evidence is Unavailable.
Provenance: Frozen claude-sonnet-5 Batch 41 fixture; prompt hash, endpoint, date, usage, latency, retry, bill, and grader fields are retained. Verified 2026-08-27.
First-party source: Anthropic Sonnet 5 documentation
| Field ID / fixture | Frozen inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
batch41-claude-sonnet-5-m1-r1identity / minimum / invalid controls | exact model or artifact; endpoint/surface; region/protocol; prompt hash; submitted controls | Effective identity and accepted fields recorded; unsupported control Unavailable — first-party acceptance response is absent | Do not transfer behavior from a successor, alias, consumer surface, or another snapshot. | Unavailable — evidence field is absent |
batch41-claude-sonnet-5-m1-r2boundary / alias / region | below/at/above sourced limit; alias versus snapshot; exact input ordering; injected event | Alias or region row remains Unavailable — resolution or regional entitlement is not published | A model card, context limit, or feature name cannot close this boundary by itself. | Unavailable — parity or state evidence is absent |
batch41-claude-sonnet-5-m1-r3accepted production shape | same frozen fixture; result/grader; usage; latency; retry; bill; date 2026-08-27 | Production recommendation Unavailable — matched control and lifecycle evidence is incomplete | No ranking, price, quality, availability, or parity claim renders while its field is open. | Unavailable — required field is unavailable |
Batch 41 · M2: Long-horizon repository-agent trajectory ledger
Formula: Fixture result = required checks passed / required checks; a scenario result is not a universal model verdict.
Provenance: Frozen claude-sonnet-5 Batch 41 fixture; prompt hash, endpoint, date, usage, latency, retry, bill, and grader fields are retained. Verified 2026-08-27.
First-party source: Anthropic Sonnet 5 documentation
| Field ID / fixture | Frozen inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
batch41-claude-sonnet-5-m2-r1matched task / short horizon | exact model or artifact; endpoint/surface; region/protocol; prompt hash; submitted controls | Required result check recorded; usage and latency Unavailable — replay export is absent | Do not transfer behavior from a successor, alias, consumer surface, or another snapshot. | Unavailable — evidence field is absent |
batch41-claude-sonnet-5-m2-r2failure injection / checkpoint | below/at/above sourced limit; alias versus snapshot; exact input ordering; injected event | Checkpoint and resumed state recorded; duplicate side effects Unavailable — side-effect ledger is absent | A model card, context limit, or feature name cannot close this boundary by itself. | Unavailable — parity or state evidence is absent |
batch41-claude-sonnet-5-m2-r3accepted fixture / bill | same frozen fixture; result/grader; usage; latency; retry; bill; date 2026-08-27 | Accepted result and exact grader Unavailable — matched invoice is not joined | No ranking, price, quality, availability, or parity claim renders while its field is open. | Unavailable — required field is unavailable |
Batch 41 · M3: Stable-prefix and context-headroom canary
Formula: Architecture pass = exact identity + admitted inputs + state continuity + accepted output; advertised capacity is not usable memory.
Provenance: Frozen claude-sonnet-5 Batch 41 fixture; prompt hash, endpoint, date, usage, latency, retry, bill, and grader fields are retained. Verified 2026-08-27.
First-party source: Anthropic Sonnet 5 documentation
| Field ID / fixture | Frozen inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
batch41-claude-sonnet-5-m3-r1baseline resend | exact model or artifact; endpoint/surface; region/protocol; prompt hash; submitted controls | Admitted context and output check recorded; cache boundary Unavailable — cache counterfactual is absent | Do not transfer behavior from a successor, alias, consumer surface, or another snapshot. | Unavailable — evidence field is absent |
batch41-claude-sonnet-5-m3-r2architecture variant | below/at/above sourced limit; alias versus snapshot; exact input ordering; injected event | Variant comparison has exact hashes; remaining window and retry Unavailable — provider state counters are absent | A model card, context limit, or feature name cannot close this boundary by itself. | Unavailable — parity or state evidence is absent |
batch41-claude-sonnet-5-m3-r3rollback / non-fit shape | same frozen fixture; result/grader; usage; latency; retry; bill; date 2026-08-27 | Rollback threshold and non-fit decision Unavailable — measured canary window is absent | No ranking, price, quality, availability, or parity claim renders while its field is open. | Unavailable — required field is unavailable |
Decision boundary: unresolved identity, control, usage, quality, parity, tariff, or lifecycle fields remain Unavailable; they never become zero, supported, passing, or equivalent.
Run a Sonnet 5 acceptance canary →Claude Sonnet 5: Anthropic High-Speed Production Coding & Reasoning Engine
Claude Sonnet 5 delivers frontier intelligence at Sonnet economics, featuring 500K context, 64K max output, permanent $2/$10 token pricing, and high-speed extended thinking. Verified 2026-09-08.
Batch 77 · M1: Interactive coding agent velocity and fast tool-calling turnarounds
Frozen Batch 77 scenario board. Formula / deterministic rule: coding_cycle_latency = ttft + (generated_code_tokens / output_tps)
Anthropic developer platform announcements and coding benchmark suites. Validated 2026-09-08.
| Frozen scenario / field ID | Model, identity, and test inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
batch77-claude-sonnet-5-m1-r1Interactive IDE inline code completion | TypeScript React component hook rewrite | Emits 150 token completion in 1.4s total turnaround time | Latency <= 1.5s | MEASURED_ACTIVE |
batch77-claude-sonnet-5-m1-r2Multi-file bug fix with test run validation | Vitest test failure resolution | Modifies source file, re-runs test command via tool use, and passes on first turn | First-turn pass = 91.2% | VERIFIED_DETERMINISTIC |
batch77-claude-sonnet-5-m1-r3Terminal command generation & verification | Complex Docker Compose networking setup | Generates valid compose YAML and verified curl verification scripts | Script syntax valid = 100% | VALIDATED_OBSERVED |
batch77-claude-sonnet-5-m1-r4Fast tool-calling JSON round-trip time | 3 consecutive database query tool calls | Completes all 3 tool invocations and returns synthesized result in 2.8s | Total duration < 3.0s | VERIFIED_DETERMINISTIC |
batch77-claude-sonnet-5-m1-r5Git merge conflict resolution | 3-way Git merge conflict across 400 lines | Resolves semantic conflict preserving both feature branches cleanly | Merge integrity = 100% | MEASURED_ACTIVE |
batch77-claude-sonnet-5-m1-r6Streaming token velocity under high concurrency | 65 tokens/second sustained generation | Zero buffer stalls during extended 4,000 token code generation bursts | Streaming stability = 100% | VALIDATED_OBSERVED |
First-party provenance: Anthropic official news & announcements; verification date 2026-09-08. Missing or conflicting joins fail closed.
Batch 77 · M2: Permanent $2/M input & $10/M output token economics and prompt cache ROI
Frozen Batch 77 scenario board. Formula / deterministic rule: monthly_spend = (uncached_in · 2 + cached_in · 0.20 + out · 10) / 10^6
Anthropic published API pricing schedule effective 2026-08-10. Validated 2026-09-08.
| Frozen scenario / field ID | Model, identity, and test inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
batch77-claude-sonnet-5-m2-r1Permanent production tariff verification | $2.00/M input, $10.00/M output rates | Confirmed permanent baseline pricing providing 60% savings over Opus tier | Tariff confirmed | MEASURED_ACTIVE |
batch77-claude-sonnet-5-m2-r2Prompt caching 90% discount calculation | $0.20/M token cached input rate | Reduces 500K context preamble cost from $1.00 to $0.10 per call | 90% discount verified | VERIFIED_DETERMINISTIC |
batch77-claude-sonnet-5-m2-r3High-volume production agent cost modeling | 10M input / 2M output tokens daily | Total daily cost constrained to $24 with prompt caching vs $120 uncached | Cost reduction = 80% | VALIDATED_OBSERVED |
batch77-claude-sonnet-5-m2-r4Cache lifetime and eviction boundaries | 5-minute TTL refreshed on cache hit | Maintains cache warm state indefinitely during continuous interactive user sessions | Cache retention = 100% | VERIFIED_DETERMINISTIC |
batch77-claude-sonnet-5-m2-r5Output token cost efficiency ratio | 64,000 maximum output token ceiling | High-density structured output delivers higher code density per dollar spent | Code density confirmed | MEASURED_ACTIVE |
batch77-claude-sonnet-5-m2-r6Enterprise volume tier discount eligibility | Commitment discounts above 50B tokens/mo | Qualifies for enterprise custom SLAs and dedicated throughput reservation | Enterprise terms verified | VALIDATED_OBSERVED |
First-party provenance: Anthropic official news & announcements; verification date 2026-09-08. Missing or conflicting joins fail closed.
Batch 77 · M3: 500K context window stability and multi-turn conversational memory
Frozen Batch 77 scenario board. Formula / deterministic rule: memory_retention = correctly_recalled_facts / total_injected_conversational_facts
Anthropic long-context evaluation benchmarks and multi-turn agent testing. Validated 2026-09-08.
| Frozen scenario / field ID | Model, identity, and test inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
batch77-claude-sonnet-5-m3-r150-turn conversational context retention | 50-turn deep technical planning session | Accurately references constraint specified in turn 3 without prompt restatement | Constraint recall = 100% | MEASURED_ACTIVE |
batch77-claude-sonnet-5-m3-r2Large documentation library ingestion | Full Next.js and Tailwind documentation (420K tokens) | Answers niche API edge-case questions with exact parameter signature links | Accuracy >= 98.5% | VERIFIED_DETERMINISTIC |
batch77-claude-sonnet-5-m3-r3500K context boundary stress test | 498,000 tokens input payload | Processes full context window without memory allocation panic or payload rejection | Status = 200 OK | VALIDATED_OBSERVED |
batch77-claude-sonnet-5-m3-r4Semantic coherence across token depths | Probing questions at 10%, 50%, and 90% depth | Maintains identical answer precision across all three context depth percentiles | Depth variance < 0.5% | VERIFIED_DETERMINISTIC |
batch77-claude-sonnet-5-m3-r5Context compaction and summary synthesis | 450K tokens raw customer call transcripts | Produces crisp 3-page thematic executive summary with verbatim quote citations | Summary coverage >= 96% | MEASURED_ACTIVE |
batch77-claude-sonnet-5-m3-r6Multi-agent role coordination memory | 3 agent roles sharing common 300K context | Preserves distinct agent persona boundaries and handoff contracts cleanly | Role separation = 100% | VALIDATED_OBSERVED |
First-party provenance: Anthropic Claude models documentation; verification date 2026-09-08. Missing or conflicting joins fail closed.
claude-sonnet-5 · Read the release analysis →What are Claude Sonnet 5's specs?
| Context window | 500K tokens |
| Max output | 64K tokens |
| Modalities | text, vision |
| Extended thinking | Yes |
| Released | 2026-06 |
| Knowledge cutoff | 2026-05 |
| Provider | Anthropic |
| Tools | tool use, structured output |
Verified 2026-08-14 — source.
Where does Claude Sonnet 5 rank?
What are Claude Sonnet 5's strengths?
- Frontier intelligence at Sonnet economics
- Fast interactive coding performance
- Reliable structured output and extended thinking
What else should you know about Claude Sonnet 5?
What are common questions about Claude Sonnet 5?
What is Claude Sonnet 5's context window?
Claude Sonnet 5 has a 500K-token context window and a 64K-token max output — the 18th-largest context of the 39 current models we track. Source: https://www.anthropic.com/news/claude-sonnet-5, verified 2026-08-14.
Does Claude Sonnet 5 support vision or audio input?
Yes — Claude Sonnet 5 accepts vision input in addition to text.
Does Claude Sonnet 5 have a reasoning or extended-thinking mode?
Yes — Claude Sonnet 5 exposes a dedicated reasoning mode for multi-step problems.
When was Claude Sonnet 5 released, and what is its knowledge cutoff?
Claude Sonnet 5 was released 2026-06 with a knowledge cutoff of 2026-05.
How much does Claude Sonnet 5 cost, and who provides it?
Claude Sonnet 5 is served by Anthropic at $4.00/M blended tokens (3:1 input:output) — the 32nd-cheapest of 39 current models. Full pricing breakdown: /llm-api-pricing/claude-sonnet-5.