Gemini 3 Flash
gemini-3-flash · Google DeepMind
Fastest model on Delphos. Optimised for speed and daily tasks — search,
summarise, draft, answer. Delivers excellent results in milliseconds, with reasoning available when
you need it.
boltFastest
visibilityVision
chatDaily tasks
searchWeb search
toggle_onToggle Reasoning
checkVision — images, screenshots,
documents
checkUltra-low latency
responses
checkToggleable reasoning / think
mode
checkWeb search & tool
execution
checkMultimodal input
support
Reasoning
toggle_on
Toggle on or off. Enable reasoning when needed — stays lightning
fast either way. Best of both worlds.
Best for
Daily conversations
Quick lookups
Summarisation
High-volume tasks
Speed-critical workflows
Gemini 3.1 Pro
gemini-3.1-pro · Google DeepMind
Google's most powerful reasoning model on Delphos. Force reasoning always
active — it thinks deeply before every response. Built for complex analysis, research, and
multi-step problems.
psychologyForce Reasoning
visibilityVision
scienceDeep analysis
searchResearch
data_objectLong context
checkVision — images, screenshots,
documents
checkForce reasoning — always
thinks before responding
checkComplex multi-step problem
solving
checkExtended context window
checkWeb search & tool
execution
Reasoning
lock
Force reasoning — always on. Cannot be disabled. Every response
includes an internal thinking pass before answering.
Best for
Deep research
Complex analysis
Multi-step tasks
Science & math
Long documents
Claude Sonnet 4.6
claude-sonnet-4-6 · Anthropic
Fast, brilliant, and deeply creative. The best all-rounder on Delphos —
exceptional at code, writing, reasoning, and instruction following. Recommended as your daily
driver.
codeBest at coding
drawCreative
visibilityVision
toggle_onToggle Reasoning
boltFast + Smart
checkVision — images, screenshots,
documents
checkToggleable reasoning / think
mode
checkBest-in-class code generation
& debugging
checkCreative writing,
storytelling, copywriting
checkInstruction following &
structured output
Reasoning
toggle_on
Toggle on or off. Enable think mode for complex problems; disable
for fast, direct answers. Full control in every conversation.
Best for
Coding & debugging
Creative writing
Daily driver
Instruction following
Balanced tasks
Claude Opus 4.6
claude-opus-4-6 · Anthropic
Anthropic's most capable model. Maximum intelligence, no compromises. For
the most demanding tasks — frontier research, complex reasoning, and nuanced long-form work.
diamondMost capable
visibilityVision
toggle_onToggle Reasoning
workspace_premiumPremium only
checkVision — images, screenshots,
documents
checkToggleable reasoning / think
mode
checkFrontier-level intelligence &
nuance
checkExtremely long context
understanding
checkComplex multi-document
analysis
Reasoning
toggle_on
Toggle on or off. Same flexible control as Sonnet — but with
Opus-level depth when thinking is enabled.
Best for
Frontier research
Complex reasoning
Long documents
Nuanced writing
Maximum quality
Llama 4 Maverick
llama-4-maverick · Meta AI
Meta's open-source frontier model. Strong, fast, and versatile — great
for general conversation, writing, and knowledge tasks. A reliable free-tier workhorse.
publicOpen source
visibilityVision
chatGeneral use
psychologyNo reasoning
checkVision — images, screenshots,
documents
checkStrong general knowledge &
conversation
checkWeb search & tool
execution
checkOpen-source, transparent
architecture
removeDeep reasoning / think
mode
Reasoning
block
No reasoning mode. Llama 4 answers directly — fast and efficient
for general-purpose conversations and tasks.
Best for
General conversation
Knowledge Q&A
Writing assistance
Free tier tasks
Open-source fans
Kimi k2 Think
kimi-k2-think · Moonshot AI
Moonshot AI's reasoning specialist. Force thinking always active — built
from the ground up to reason through hard problems. Excellent for math, logic, and multi-step
inference.
psychologyForce Thinking
calculateMath & Logic
account_treeMulti-step
hide_imageNo vision
removeNo vision — text-only
input
checkForce thinking — always
reasons before answering
checkAdvanced math, logic, and
reasoning chains
checkLong chain-of-thought
processing
checkComplex inference & deduction
tasks
Reasoning
lock
Force thinking — always on. Kimi k2 always engages its full
reasoning chain. Cannot be turned off — every answer is thought through deeply.
Best for
Mathematics
Logic problems
Chain-of-thought
Scientific reasoning
Hard inference