February
Major Release
Google's flagship reasoning model with a 2x jump on hard multi-step tasks. — Google's latest flagship model with a major 2x jump in reasoning capabilities
💰 $2.50 in / $10.00 out per 1M tok
🎛 In: text, image, audio, video, PDF
📤 Out: text
🌐 Google AI Studio · Vertex AI · Gemini API
✨ Key Features
- 2x reasoning improvement
- ARC-AGI-2 score of 77.1%
- Enhanced multimodal understanding
- Deep Think mode
🔄 What's new vs previous version
- 2x reasoning score on ARC-AGI-2 vs Gemini 3 Pro
- Context window expanded to 2M tokens
- Deep Think mode enabled by default on the Pro tier
- Lower latency on first-token despite larger context
Major Release
Near-Opus quality at a fraction of the cost, with Agent Teams orchestration. — Anthropic's latest Sonnet with Agent Teams capability and near-Opus performance at a fraction of the cost
💰 $3.00 in / $15.00 out per 1M tok
🎛 In: text, image, PDF
📤 Out: text
🌐 Anthropic API · AWS Bedrock · Google Vertex AI
✨ Key Features
- Agent Teams: orchestrate 2-16 Claude instances
- Near-Opus performance at 1/5th cost
- 80.8% SWE-bench Verified
- Fast mode research preview
🔄 What's new vs previous version
- Agent Teams: orchestrate 2–16 Claude instances in parallel
- +8.5pt on SWE-bench Verified vs Sonnet 4
- 1/5 the cost of Opus 4.5 at ~95% of coding quality
- Fast mode research preview for lower-latency inference
Update
Open-weight MoE with a 1M+ token context window and strong coding. — Major update with 10x context window expansion to over 1 million tokens
💰 $0.27 in / $1.10 out per 1M tok
🎛 In: text
📤 Out: text
🌐 DeepSeek API · Hugging Face · Together AI · Fireworks AI
✨ Key Features
- 1M+ token context window (10x expansion)
- Improved reasoning capabilities
- Open source release
- Cost-effective inference
🔄 What's new vs previous version
- 10x context window expansion (128K → 1M+ tokens)
- Sliding-window attention for long-context throughput
- Improved chain-of-thought reasoning
- Native FP8 inference support
Major Release
First frontier model trained entirely on Huawei Ascend silicon. — First frontier AI model trained entirely without NVIDIA GPUs, using Huawei Ascend chips
💰 $0.11 in / $0.28 out per 1M tok
🎛 In: text, image
📤 Out: text
🌐 Zhipu BigModel API
✨ Key Features
- First frontier model trained on Huawei Ascend chips (no NVIDIA)
- #1 HLE score (50.4%)
- 1.2% hallucination rate via Slime RL
- 136x cheaper than Claude Opus 4.5
🔄 What's new vs previous version
- Trained entirely on Huawei Ascend 910B clusters (no NVIDIA)
- Slime RL fine-tuning drops hallucination rate to 1.2%
- 136x cheaper than Claude Opus 4.5 at comparable quality
Major Release
ByteDance's video generation flagship — cinematic quality at scale — First commercial video model generating synchronized audio and video in a single pass
🎛 In: text, image
📤 Out: video
🌐 Volcengine API
✨ Key Features
- Synchronized audio + video generation in one pass
- Lip-sync in 8+ languages
- 10-30x cheaper than Sora 2
- Commercial pricing: $0.10-$0.80/min
Major Release
Coding-specialized variant of GPT-5.3, tuned for agentic IDE workflows. — OpenAI's specialized self-improving coding model with state-of-the-art software engineering performance
💰 $1.25 in / $10.00 out per 1M tok
🎛 In: text, image
📤 Out: text
🌐 OpenAI API · Azure OpenAI · GitHub Copilot
✨ Key Features
- Self-improving agentic coding
- 25% faster than GPT-5.2-Codex
- 1,000+ tokens/sec generation
- First OpenAI model flagged 'high' on cybersecurity framework
🔄 What's new vs previous version
- +4pt on SWE-bench Verified vs GPT-5.2 Codex
- Native IDE tool-calling at reduced latency
- Extended max output to 100K for multi-file patches