Amazon Nova 2 Lite surpasses Nova 1 Pro with 1M token context and extended thinking at 7x lower cost
Amazon Nova 2 Lite expands context window to 1 million tokens, introduces extended thinking with developer controls, and adds native tool use and web grounding. AWS claims Nova 2 Lite surpasses Nova 1 Pro on multi-step reasoning while costing 7x less and running up to 5x faster.
Amazon Nova 2 Lite — Quick Specs
Amazon Nova 2 Lite Replaces Nova 1 Across All Tiers With Major Capability Expansion
Amazon has released detailed migration guidance for moving from Amazon Nova 1 to Nova 2 on Bedrock, positioning Nova 2 Lite as a direct replacement across all three Nova 1 tiers despite the apparent tier downgrade.
What Changed: Context and Reasoning
Nova 2 Lite expands the context window from 300K to 1M tokens—a 3.3x increase—and raises maximum output from 10K to 65K tokens. The model adds extended thinking with developer controls (low, medium, high reasoning effort levels), native tool use with support for MCP servers and parallel tool chaining, built-in web grounding with citations, and a Python code interpreter.
All Nova 2 features maintain input modality support for text, image, and video, matching Nova 1's multimodal capabilities.
Migration Paths and Performance Claims
AWS recommends Nova 2 Lite as the upgrade path for all three Nova 1 tiers:
Nova 1 Lite → Nova 2 Lite: Straightforward upgrade maintaining the same capabilities plus extended thinking and 1M context.
Nova 1 Pro → Nova 2 Lite: AWS recommends this apparent tier downgrade, claiming Nova 2 Lite with extended thinking handles workloads previously requiring Pro while delivering 7x lower cost and up to 5x faster inference on multi-step problem-solving.
Nova 1 Premier → Nova 2 Lite: For agentic and tool use workloads, AWS suggests Nova 2 Lite remains cheaper and faster than Premier, recommending evaluation across reasoning effort levels to verify quality.
Benchmark Scores
Nova 2 Lite achieves:
- 80.9% on MMLU Pro
- 70.8% on IF-Bench
- 76.0% on τ2-bench Telecom (tool calling benchmark)
AWS provides no direct benchmark comparisons between Nova 1 Pro/Premier and Nova 2 Lite in the migration guide, though the technical report claims superiority on "multi-step problem-solving."
Customer Deployments
AWS highlights three production use cases:
Siemens Global Search: Claims 300% search speed improvement and 70% cost reduction versus previous LLM solution running on Nova 2 Lite.
Trellix Security Alert Triage: Reports 39% accuracy improvement in threat classification and 3.4x more detailed responses with tool calling, with zero tool calling failures after migration.
AWS Transform: Multi-agent infrastructure modernization system claims up to 60% improvement in tool calling efficiency for code modernization.
Recommended Configuration
AWS suggests:
- Start multi-step agentic workflows with reasoning set to "Low"
- Evaluate quality before moving to medium or high reasoning effort
- For Pro/Premier migrations, test with extended thinking enabled to verify quality on existing workloads
What This Means
Amazon is consolidating its Nova lineup around a single production-ready tier (Nova 2 Lite) while expanding capabilities that previously required larger models. The 7x cost reduction claim and 5x speed advantage on reasoning tasks, if accurate, represent significant efficiency gains for enterprise customers. However, independent benchmarking comparing Nova 2 Lite directly to Nova 1 Pro/Premier on the same tasks is absent—AWS relies on technical report claims rather than transparent benchmark tables.
The strategy prioritizes agentic AI, tool use, and document processing workloads. Extended thinking with developer controls suggests AWS is implementing reasoning tokens similar to o1-style models, allowing cost/quality tradeoffs rather than fixed inference behavior. Pricing details remain undisclosed in this migration guide.
Related Articles
Google rolls out Personal Intelligence to all Gemini users, accessing Gmail and search history
Google has expanded Personal Intelligence, its hyper-personalized Gemini mode, from $20/month subscribers to all users. The feature integrates data from Gmail, Search history, Google Photos, and other Google services to provide contextual assistance, though it remains entirely opt-in.
Midjourney V8 achieves 5x faster generation but premium features cost 4x more
Midjourney has released an early version of V8 for community testing, achieving roughly 5x faster image generation and introducing native 2K resolution via --hd mode. However, premium features including --hd, --q 4, style references, and mood boards cost four times as much as standard generation, with Relax mode unavailable at launch.
Meta's Manus launches desktop app enabling AI agents to access local files and applications
Meta's recently acquired AI startup Manus launched a desktop application enabling its AI agent to directly access local files, tools, and applications on personal computers through a 'My Computer' feature. Previously cloud-only, the move positions Manus to compete with OpenClaw, the open-source AI agent that sparked recent industry momentum. Unlike OpenClaw's free, MIT-licensed offering, Manus operates as a paid subscription service.
DuckDuckGo adds GPT-5 mini and GPT-5.2 reasoning models to Duck.ai privacy chatbot
DuckDuckGo's Duck.ai chatbot platform now includes OpenAI's GPT-5 mini for free users and GPT-5.2 for subscribers, both with reasoning capabilities. The platform continues to anonymize all conversations by default, stripping metadata before routing chats to model providers including Anthropic, Meta, Mistral, and OpenAI.