Skip to content

Providers

The catalogue of LLM providers shipped with LazyBridge, the tier aliases each one resolves, and the per-provider quirks (thinking modes, native tools, deprecation timelines). For writing a brand-new provider see BaseProvider.

Pricing and model lineup snapshot from July 2026. LLM provider economics shift fast — treat the tables below as a structural reference (which alias resolves to which model, which features work on which model) rather than as live pricing.

Signature

from lazybridge import Agent, LLMEngine

# Direct model selection — provider inferred from the model string.
Agent(engine=LLMEngine("claude-opus-5"))
Agent(engine=LLMEngine("gpt-5.6-luna"))

# Tier-based selection — model never appears in app code.
Agent.from_provider("anthropic", tier="top")     # → claude-fable-5
Agent.from_provider("openai",    tier="medium")  # → gpt-5.6-luna
Agent.from_provider("google",    tier="cheap")   # → gemini-3.1-flash-lite-preview

Agent.from_provider is sugar for Agent(engine=LLMEngine(<resolved-model>, provider=<name>)). See Canonical vs sugar for the breakdown.

Tier names

Tier Intent
super_cheap Smallest / cheapest model in the lineup; for parsing, classification, throwaway calls
cheap Default budget tier
medium The default for Agent.from_provider(...)
expensive Premium reasoning / long-context tier
top The flagship model

Each provider's _TIER_ALIASES table maps these strings to a concrete model name. A string not in the table is treated as a literal model name (passthrough).

Built-in providers

Anthropic

tier model ctx max_out $/M in $/M out
top claude-fable-5 1 M 128 K $10.00 $50.00
expensive claude-opus-5 1 M 128 K $5.00 $25.00
medium claude-sonnet-5 1 M 128 K $2.00¹ $10.00¹
cheap claude-haiku-4-5 200 K 64 K $1.00 $5.00
super_cheap claude-3-haiku 200 K 4 K $0.25 $1.25

¹ Sonnet 5 introductory pricing through 2026-08-31; rises to $3.00 / $15.00 per million tokens after that date.

Not tier-aliased: claude-mythos-5 ($10.00 / $50.00, same underlying model as Fable 5 with fewer safety guardrails) — restricted to vetted partners (Project Glasswing / US government cyber defenders), not reachable with an ordinary API key. Older pinned ids (claude-opus-4-8, claude-opus-4-7, claude-opus-4-6, claude-sonnet-4-6, claude-opus-4-1, …) still resolve — see _PRICE_TABLE in core/providers/anthropic.py for the full list.

  • Thinking. fable-5 / mythos-5 / opus-5 / opus-4-8 / opus-4-7 / opus-4-6 / sonnet-5 / sonnet-4-6 use adaptive thinking (no budget_tokens argument). haiku-4-5 and earlier 3.x models require ThinkingConfig(budget_tokens=N). fable-5 / mythos-5 / opus-5 / opus-4-8 / opus-4-7 / sonnet-5 do not accept temperature / top_p / top_k.
  • Effort. fable-5, mythos-5, opus-5, opus-4-8, opus-4-7, opus-4-6, sonnet-5, sonnet-4-6, and opus-4-5 accept ThinkingConfig(effort=...) ∈ {low, medium, high, xhigh, max} (default high), sent as output_config.effort. It works with or without thinking enabled and, unlike budget_tokens, shapes all response tokens (text, tool calls, thinking). xhigh isn't available on opus-4-6 / sonnet-4-6 / opus-4-5 — LazyBridge downgrades an xhigh request on those models to max with a warning. Shorthand: LLMEngine(model, thinking="low").
  • Native tools. WEB_SEARCH, CODE_EXECUTION, COMPUTER_USE.

OpenAI

tier model ctx max_out $/M in $/M cached $/M out
top gpt-5.6-sol 1.05 M 128 K $5.00 $0.50 $30.00
expensive gpt-5.6-terra 1.05 M 128 K $2.50 $0.25 $15.00
medium gpt-5.6-luna 1.05 M 128 K $1.00 $0.10 $6.00
cheap gpt-5.4-nano 400 K 128 K $0.20 $0.02 $1.25
super_cheap gpt-4o-mini 128 K 16 K $0.15 — $0.60

GPT-5.6 (released 2026-07-09) replaced the old flagship+-pro shape with three tiers: Sol (best coding / hardest reasoning, OpenAI's "workhorse"), Terra (balanced general flagship), Luna (fast/light). The bare alias gpt-5.6 routes to Sol. GPT-5.6 also introduces explicit prompt-cache breakpoints and a 30-minute minimum cache life.

Other supported models (passed verbatim, no tier alias): gpt-5.5-pro ($30 / $180), gpt-5.5 ($5 / $0.50 cache / $30), gpt-5.4-pro ($30 / $180), gpt-5.4 ($2.50 / $0.25 cache / $15), gpt-5.4-mini ($0.75 / $0.075 cache / $4.50), gpt-5 ($1.25 / $10), gpt-4o ($2.50 / $10), gpt-4.1 ($2 / $8), gpt-4.1-mini ($0.40 / $1.60), o3 ($2 / $8), o4-mini ($1.10 / $4.40).

  • Thinking. gpt-5.6-* / gpt-5.5 / gpt-5.5-pro accept reasoning_effort ∈ {none, low, medium, high, xhigh} (default medium). The o-series and gpt-5.4-pro accept reasoning_effort ∈ {low, medium, high}. Standard GPT models don't support thinking. Shorthand: LLMEngine(model, thinking="low") — mirrors the Anthropic ThinkingConfig.effort shorthand, since both providers read the same ThinkingConfig.effort field.
  • Native tools. WEB_SEARCH, CODE_EXECUTION, FILE_SEARCH, COMPUTER_USE, IMAGE_GENERATION.
  • Cache. Automatic via prompt_tokens_details.cached_tokens; cached_input rate applied when published (gpt-5.6-*, gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.4-nano).
  • Long-context surcharge (>272K input on gpt-5.x) is not modeled in cost rollup — the reported cost may under-count for large prompts.
  • Not modeled. OpenAI's realtime voice models (GPT-Live-1, GPT-Live-1 mini, released 2026-07-08) are a separate full-duplex Realtime API, not the text Responses/Chat Completions path OpenAIProvider wraps — out of scope for this provider.

Google

tier model ctx max_out $/M in $/M out
top gemini-3.1-pro-preview 1 M 64 K $2.00 $12.00
expensive gemini-2.5-pro 1 M 64 K $1.25 $10.00
medium gemini-3-flash-preview 1 M 64 K $0.50 $3.00
cheap gemini-3.1-flash-lite-preview 1 M 64 K $0.25 $1.50
super_cheap gemini-2.5-flash-lite 1 M 64 K $0.10 $0.40
  • Thinking. gemini-3.x accepts ThinkingConfig(thinking_level=...) with low / medium / high. gemini-2.x accepts ThinkingConfig(thinking_budget=N); -1 selects auto-budget.
  • Native tools. GOOGLE_SEARCH, WEB_SEARCH, GOOGLE_MAPS.
  • Warning. Google Search + structured output produces a provider 400 — they're mutually exclusive.
  • Deprecation. gemini-2.0-flash retires June 1 2026; do not use in new code.

DeepSeek

Prices below are peak-hour rates (01:00-04:00 and 06:00-10:00 UTC); LazyBridge always costs at the peak rate for a conservative estimate. Off-peak rates are exactly half.

tier model ctx max_out $/M in $/M cached $/M out
top / expensive deepseek-v4-pro 1 M 384 K $1.32 $0.044 $3.96
medium / cheap / super_cheap deepseek-v4-flash 1 M 384 K $0.44 $0.014 $1.32
  • Thinking. Both V4 models accept ThinkingConfig → reasoning_content field on the response. In thinking mode the provider strips temperature / top_p / presence_penalty / frequency_penalty. ThinkingConfig on non-V4 models raises ValueError.
  • Cache. Automatic on repeated prefixes ≥1024 tokens; no opt-in required.
  • Native tools. None (function calling is supported).
  • Deprecation (retire 2026-07-24). deepseek-reasoner and deepseek-chat both alias to deepseek-v4-flash.

LMStudio

A local OpenAI-compatible runtime. LMStudioProvider extends OpenAIProvider; point OPENAI_BASE_URL at your LM Studio instance and use any model name your local install serves.

LiteLLM

The unified bridge for the long tail (Mistral, Cohere, Groq, Bedrock, Vertex, Ollama, etc.). Use the litellm/ model-string prefix to route through LiteLLMProvider. Native providers (Anthropic, OpenAI, Google, DeepSeek) still handle their own models directly — LiteLLM is the catch-all for the rest.

Agent(engine=LLMEngine("litellm/groq/llama-3.3-70b"))

tool_choice values

LLMEngine accepts a tool_choice= kwarg that drives provider tool selection:

Value Meaning
"auto" Model decides (default)
"none" No tool calls allowed
"required" Must call at least one tool
"any" Alias for "required"; mapped to provider equivalent ("required" for OpenAI, {"type":"required"} for Anthropic)
"<tool_name>" Must call the named tool

After the first tool-call turn, tool_choice resets to "auto" automatically — so a forced first invocation doesn't lock the rest of the loop.

DeepSeek does not support tool_choice in thinking mode.

Google finish_reason mapping

The Google provider normalises finish_reason strings so callers don't have to switch on Gemini-specific values:

Gemini value Normalised
MAX_TOKENS "max_tokens"
SAFETY / RECITATION / BLOCKLIST / PROHIBITED_CONTENT / SPII "stop"
anything else "end_turn"

Pitfalls

  • DeepSeek tier collapse. Three of the five tier aliases (medium / cheap / super_cheap) all map to deepseek-v4-flash — there's no smaller model in the lineup.
  • gpt-5.6-nano doesn't exist. The cheap tier stays on gpt-5.4-nano — Luna is the fast/light GPT-5.6 tier but isn't actually cheaper per-token than 5.4-nano.
  • gemini-2.0-flash deprecation lands June 1 2026; switch to gemini-2.5-flash-lite before then.
  • Adaptive thinking ignores budget_tokens. Anthropic claude-fable/mythos/opus 5, claude-opus / claude-sonnet 4.6+ pick their own thinking budget; passing ThinkingConfig(budget_tokens=...) is no-effect. Use ThinkingConfig(effort=...) instead.
  • claude-mythos-5 is access-restricted, not a public tier — don't wire it into Agent.from_provider("anthropic", tier=...).
  • tool_choice="any" is not passed literally. It maps to "required" (or the provider equivalent) at request time.
  • Pricing changes faster than these tables. Check the provider's current rate card before reasoning about cost in production.

See also

  • BaseProvider — write your own provider when none of the built-ins fits.
  • Native tools — what each provider exposes server-side; the per-provider table above lists the supported NativeTool enum values.
  • Canonical vs sugar — Agent.from_provider("…", tier="top") is one of the few factory methods that's not pure sugar (it builds the engine with the tier alias and an explicit provider=).