Skip to content

Anthropic

Anthropic is a first-class execution provider on LenserFight. The platform supports the Claude 4.x Opus and Sonnet tiers for demanding reasoning and agentic workloads, as well as the Haiku 4.5 and 3.5 models for high-throughput, cost-sensitive lenses. All Claude models on LenserFight share the 200 000-token context window.

Support tier

runnable — LenserFight has a direct runtime path or adapter for this provider.

Upstream docs

docs.anthropic.com

Models on LenserFight

Claude Opus 4.6

FieldValue
Keyclaude-opus-4-6
Capabilitieschat · reasoning · tools
Context window200 000 tokens
Input modalitiestext
Output modalitiestext
Provider docs

Anthropic's most capable model. Suited for long-document analysis, multi-step reasoning chains, and complex agentic tasks.

Claude Sonnet 4.6

FieldValue
Keyclaude-sonnet-4-6
Capabilitieschat · reasoning · tools
Context window200 000 tokens
Input modalitiestext
Output modalitiestext
Provider docs

Balanced speed/quality. Default recommendation for most Claude-backed lenses.

Claude Sonnet 4.5

FieldValue
Keyclaude-sonnet-4-5
Capabilitieschat · reasoning · tools
Context window200 000 tokens
Input modalitiestext
Output modalitiestext
Provider docs

Claude Sonnet 4.0

FieldValue
Keyclaude-sonnet-4-0
Capabilitieschat · reasoning · tools
Context window200 000 tokens
Input modalitiestext
Output modalitiestext
Provider docs

Claude Haiku 4.5

FieldValue
Keyclaude-haiku-4-5
Capabilitieschat
Context window200 000 tokens
Input modalitiestext
Output modalitiestext
Provider docs

Fastest, cheapest Anthropic model. Use for high-volume classification or summarisation where speed matters most.

Claude Haiku 3.5

FieldValue
Keyclaude-haiku-3-5
Capabilitieschat
Context window200 000 tokens
Input modalitiestext
Output modalitiestext
Provider docs

Legacy. Prefer Haiku 4.5 for new lenses.

Usage notes

  • Claude models on LenserFight are text-only; they do not accept image or document inputs in the default catalog configuration.
  • For tool-calling lenses, Claude Sonnet 4.6 offers the best latency-to-capability trade-off; reserve Opus 4.6 for tasks that genuinely require extended reasoning chains.