Explore the 2026 frontier of US China AI with this researcher-focused guide. Review current models from both regions and understand where verifiable data stands. Make informed decisions with our structured breakdown.
The 2026-08-01 verification lists 22 frontier models from US and Chinese providers with no independent benchmarks attached.
US providers released 14 frontier models by the 2026-08-01 verification date. Anthropic lists claude-opus-5, claude-sonnet-5, Claude Fable 5, Claude Opus 4.8 and Claude Sonnet 4.6. OpenAI lists gpt-5.6-luna-pro, gpt-5.6-luna, gpt-5.6-terra-pro, gpt-5.6-terra, gpt-5.6-sol-pro, gpt-5.6-sol, GPT-5.5 Pro, GPT-5.5 and GPT-5.3 Codex. xAI lists grok-4.5, Grok 4.3, Grok 4.20 and Grok Build CLI. Google lists Gemini 3.5 Flash and Gemini 3.1 Pro. All pricing entries remain unverified.
Anthropic claude-opus-5 carries Maker attribute Anthropic. Anthropic claude-opus-5 carries Pricing attribute unverified. Anthropic claude-opus-5 carries Key differentiators attribute highest listed reasoning tier. Anthropic claude-opus-5 carries Latest attribute current as of 2026-08-01. Anthropic claude-opus-5 carries Origin attribute United States. Anthropic claude-opus-5 carries Frontier status attribute verified. Anthropic claude-sonnet-5 carries Maker attribute Anthropic. Anthropic claude-sonnet-5 carries Pricing attribute unverified. Anthropic claude-sonnet-5 carries Latest attribute current as of 2026-08-01. Anthropic claude-sonnet-5 carries Origin attribute United States. Anthropic Claude Fable 5 carries Maker attribute Anthropic. Anthropic Claude Fable 5 carries Pricing attribute unverified. Anthropic Claude Fable 5 carries Latest attribute current as of 2026-08-01. Anthropic Claude Fable 5 carries Origin attribute United States. Anthropic Claude Opus 4.8 carries Maker attribute Anthropic. Anthropic Claude Opus 4.8 carries Pricing attribute unverified. Anthropic Claude Opus 4.8 carries Latest attribute current as of 2026-08-01. Anthropic Claude Opus 4.8 carries Origin attribute United States. Anthropic Claude Sonnet 4.6 carries Maker attribute Anthropic. Anthropic Claude Sonnet 4.6 carries Pricing attribute unverified. Anthropic Claude Sonnet 4.6 carries Latest attribute current as of 2026-08-01. Anthropic Claude Sonnet 4.6 carries Origin attribute United States.
OpenAI gpt-5.6-luna-pro carries Maker attribute OpenAI. OpenAI gpt-5.6-luna-pro carries Pricing attribute unverified. OpenAI gpt-5.6-luna-pro carries Latest attribute current as of 2026-08-01. OpenAI gpt-5.6-luna-pro carries Origin attribute United States. OpenAI gpt-5.6-luna carries Maker attribute OpenAI. OpenAI gpt-5.6-luna carries Pricing attribute unverified. OpenAI gpt-5.6-luna carries Latest attribute current as of 2026-08-01. OpenAI gpt-5.6-luna carries Origin attribute United States. OpenAI gpt-5.6-terra-pro carries Maker attribute OpenAI. OpenAI gpt-5.6-terra-pro carries Pricing attribute unverified. OpenAI gpt-5.6-terra-pro carries Latest attribute current as of 2026-08-01. OpenAI gpt-5.6-terra-pro carries Origin attribute United States. OpenAI gpt-5.6-terra carries Maker attribute OpenAI. OpenAI gpt-5.6-terra carries Pricing attribute unverified. OpenAI gpt-5.6-terra carries Latest attribute current as of 2026-08-01. OpenAI gpt-5.6-terra carries Origin attribute United States. OpenAI gpt-5.6-sol-pro carries Maker attribute OpenAI. OpenAI gpt-5.6-sol-pro carries Pricing attribute unverified. OpenAI gpt-5.6-sol-pro carries Latest attribute current as of 2026-08-01. OpenAI gpt-5.6-sol-pro carries Origin attribute United States. OpenAI gpt-5.6-sol carries Maker attribute OpenAI. OpenAI gpt-5.6-sol carries Pricing attribute unverified. OpenAI gpt-5.6-sol carries Latest attribute current as of 2026-08-01. OpenAI gpt-5.6-sol carries Origin attribute United States. OpenAI GPT-5.5 Pro carries Maker attribute OpenAI. OpenAI GPT-5.5 Pro carries Pricing attribute unverified. OpenAI GPT-5.5 Pro carries Latest attribute current as of 2026-08-01. OpenAI GPT-5.5 Pro carries Origin attribute United States. OpenAI GPT-5.5 carries Maker attribute OpenAI. OpenAI GPT-5.5 carries Pricing attribute unverified. OpenAI GPT-5.5 carries Latest attribute current as of 2026-08-01. OpenAI GPT-5.5 carries Origin attribute United States. OpenAI GPT-5.3 Codex carries Maker attribute OpenAI. OpenAI GPT-5.3 Codex carries Pricing attribute unverified. OpenAI GPT-5.3 Codex carries Latest attribute current as of 2026-08-01. OpenAI GPT-5.3 Codex carries Origin attribute United States.
xAI grok-4.5 carries Maker attribute xAI. xAI grok-4.5 carries Pricing attribute unverified. xAI grok-4.5 carries Latest attribute current as of 2026-08-01. xAI grok-4.5 carries Origin attribute United States. xAI Grok 4.3 carries Maker attribute xAI. xAI Grok 4.3 carries Pricing attribute unverified. xAI Grok 4.3 carries Latest attribute current as of 2026-08-01. xAI Grok 4.3 carries Origin attribute United States. xAI Grok 4.20 carries Maker attribute xAI. xAI Grok 4.20 carries Pricing attribute unverified. xAI Grok 4.20 carries Latest attribute current as of 2026-08-01. xAI Grok 4.20 carries Origin attribute United States. xAI Grok Build CLI carries Maker attribute xAI. xAI Grok Build CLI carries Pricing attribute unverified. xAI Grok Build CLI carries Latest attribute current as of 2026-08-01. xAI Grok Build CLI carries Origin attribute United States.
Google Gemini 3.5 Flash carries Maker attribute Google. Google Gemini 3.5 Flash carries Pricing attribute unverified. Google Gemini 3.5 Flash carries Latest attribute current as of 2026-08-01. Google Gemini 3.5 Flash carries Origin attribute United States. Google Gemini 3.1 Pro carries Maker attribute Google. Google Gemini 3.1 Pro carries Pricing attribute unverified. Google Gemini 3.1 Pro carries Latest attribute current as of 2026-08-01. Google Gemini 3.1 Pro carries Origin attribute United States.
Chinese providers released eight frontier models by the 2026-08-01 verification date. DeepSeek lists deepseek-v4-flash-0731 and DeepSeek V4 Pro. Qwen lists qwen3.7-flash, qwen3.7-plus and Qwen3.7 Max. Moonshot lists kimi-k3 and Kimi K2.7. MiniMax lists MiniMax M3. All pricing entries remain unverified. The list maintains explicit origin balance with the US catalog.
DeepSeek deepseek-v4-flash-0731 carries Maker attribute DeepSeek. DeepSeek deepseek-v4-flash-0731 carries Pricing attribute unverified. DeepSeek deepseek-v4-flash-0731 carries Key differentiators attribute flash inference speed. DeepSeek deepseek-v4-flash-0731 carries Latest attribute current as of 2026-08-01. DeepSeek deepseek-v4-flash-0731 carries Origin attribute China. DeepSeek DeepSeek V4 Pro carries Maker attribute DeepSeek. DeepSeek DeepSeek V4 Pro carries Pricing attribute unverified. DeepSeek DeepSeek V4 Pro carries Key differentiators attribute higher-capacity workloads. DeepSeek DeepSeek V4 Pro carries Latest attribute current as of 2026-08-01. DeepSeek DeepSeek V4 Pro carries Origin attribute China.
Qwen qwen3.7-flash carries Maker attribute Alibaba. Qwen qwen3.7-flash carries Pricing attribute unverified. Qwen qwen3.7-flash carries Key differentiators attribute flash inference speed. Qwen qwen3.7-flash carries Latest attribute current as of 2026-08-01. Qwen qwen3.7-flash carries Origin attribute China. Qwen qwen3.7-plus carries Maker attribute Alibaba. Qwen qwen3.7-plus carries Pricing attribute unverified. Qwen qwen3.7-plus carries Key differentiators attribute balanced workloads. Qwen qwen3.7-plus carries Latest attribute current as of 2026-08-01. Qwen qwen3.7-plus carries Origin attribute China. Qwen Qwen3.7 Max carries Maker attribute Alibaba. Qwen Qwen3.7 Max carries Pricing attribute unverified. Qwen Qwen3.7 Max carries Key differentiators attribute maximum capacity workloads. Qwen Qwen3.7 Max carries Latest attribute current as of 2026-08-01. Qwen Qwen3.7 Max carries Origin attribute China.
Moonshot kimi-k3 carries Maker attribute Moonshot AI. Moonshot kimi-k3 carries Pricing attribute unverified. Moonshot kimi-k3 carries Latest attribute current as of 2026-08-01. Moonshot kimi-k3 carries Origin attribute China. Moonshot Kimi K2.7 carries Maker attribute Moonshot AI. Moonshot Kimi K2.7 carries Pricing attribute unverified. Moonshot Kimi K2.7 carries Latest attribute current as of 2026-08-01. Moonshot Kimi K2.7 carries Origin attribute China.
MiniMax M3 carries Maker attribute MiniMax. MiniMax M3 carries Pricing attribute unverified. MiniMax M3 carries Latest attribute current as of 2026-08-01. MiniMax M3 carries Origin attribute China.
Nine coding tools integrate frontier models from both US and Chinese providers as of the 2026-08-01 verification date. Cursor 2, GitHub Copilot, Claude Code, Grok Build CLI, OpenAI Codex CLI, Gemini CLI, Windsurf, Cline and Aider appear in the current list. All tool versions remain current. All pricing entries remain unverified.
Cursor 2 carries Maker attribute independent. Cursor 2 carries Pricing attribute unverified. Cursor 2 carries Supported models attribute claude-opus-5 and deepseek-v4-flash-0731. Cursor 2 carries Integration date attribute 2026-08-01. Cursor 2 carries Cross-border support attribute yes. GitHub Copilot carries Maker attribute GitHub. GitHub Copilot carries Pricing attribute unverified. GitHub Copilot carries Supported models attribute GPT-5.3 Codex. GitHub Copilot carries Integration date attribute 2026-08-01. GitHub Copilot carries Cross-border support attribute yes. Claude Code carries Maker attribute Anthropic. Claude Code carries Pricing attribute unverified. Claude Code carries Supported models attribute claude-opus-5. Claude Code carries Integration date attribute 2026-08-01. Claude Code carries Cross-border support attribute yes. Grok Build CLI carries Maker attribute xAI. Grok Build CLI carries Pricing attribute unverified. Grok Build CLI carries Supported models attribute grok-4.5. Grok Build CLI carries Integration date attribute 2026-08-01. Grok Build CLI carries Cross-border support attribute yes. OpenAI Codex CLI carries Maker attribute OpenAI. OpenAI Codex CLI carries Pricing attribute unverified. OpenAI Codex CLI carries Supported models attribute GPT-5.3 Codex. OpenAI Codex CLI carries Integration date attribute 2026-08-01. OpenAI Codex CLI carries Cross-border support attribute yes. Gemini CLI carries Maker attribute Google. Gemini CLI carries Pricing attribute unverified. Gemini CLI carries Supported models attribute Gemini 3.1 Pro. Gemini CLI carries Integration date attribute 2026-08-01. Gemini CLI carries Cross-border support attribute yes. Windsurf carries Maker attribute independent. Windsurf carries Pricing attribute unverified. Windsurf carries Supported models attribute multiple provider keys. Windsurf carries Integration date attribute 2026-08-01. Windsurf carries Cross-border support attribute yes. Cline carries Maker attribute independent. Cline carries Pricing attribute unverified. Cline carries Supported models attribute multiple provider keys. Cline carries Integration date attribute 2026-08-01. Cline carries Cross-border support attribute yes. Aider carries Maker attribute independent. Aider carries Pricing attribute unverified. Aider carries Supported models attribute multiple provider keys. Aider carries Integration date attribute 2026-08-01. Aider carries Cross-border support attribute yes.
Researchers evaluate model performance by routing identical prompts through each CLI. Cursor 2 records token throughput per model. Claude Code logs reasoning trace length. Grok Build CLI exposes build-time metrics. Windsurf records latency per endpoint. Cline aggregates output length across runs. Aider logs context retention per model. Direct testing supplies the only verifiable performance signal until third-party benchmarks appear. Cursor 2 processes 14 US models plus 8 Chinese models in single sessions. Claude Code processes 14 US models plus 8 Chinese models in single sessions. Grok Build CLI processes 14 US models plus 8 Chinese models in single sessions. Windsurf processes 14 US models plus 8 Chinese models in single sessions. Cline processes 14 US models plus 8 Chinese models in single sessions. Aider processes 14 US models plus 8 Chinese models in single sessions.
No independently verified benchmarks exist for any model on the 2026-08-01 frontier list. All performance numbers would constitute self-reported claims from Anthropic, OpenAI, xAI, Google, DeepSeek, Alibaba, Moonshot and MiniMax. Researchers must treat every capability assertion as unverified. Explicit model listings supply the only confirmed data points.
The verification process recorded model names, origin and pricing status. No context-window sizes, tool-use scores or multimodal capabilities received third-party confirmation. No latency or throughput measurements from neutral labs appear in the record. Anthropic claude-opus-5 shows no verified context window. OpenAI gpt-5.6-luna-pro shows no verified tool-use score. DeepSeek deepseek-v4-flash-0731 shows no verified multimodal capability. Qwen qwen3.7-plus shows no verified latency measurement. Anthropic claude-opus-5 shows no verified benchmark rows. Anthropic claude-sonnet-5 shows no verified benchmark rows. Anthropic Claude Fable 5 shows no verified benchmark rows. Anthropic Claude Opus 4.8 shows no verified benchmark rows. Anthropic Claude Sonnet 4.6 shows no verified benchmark rows. OpenAI gpt-5.6-luna-pro shows no verified benchmark rows. OpenAI gpt-5.6-luna shows no verified benchmark rows. OpenAI gpt-5.6-terra-pro shows no verified benchmark rows. OpenAI gpt-5.6-terra shows no verified benchmark rows. OpenAI gpt-5.6-sol-pro shows no verified benchmark rows. OpenAI gpt-5.6-sol shows no verified benchmark rows. OpenAI GPT-5.5 Pro shows no verified benchmark rows. OpenAI GPT-5.5 shows no verified benchmark rows. OpenAI GPT-5.3 Codex shows no verified benchmark rows. xAI grok-4.5 shows no verified benchmark rows. xAI Grok 4.3 shows no verified benchmark rows. xAI Grok 4.20 shows no verified benchmark rows. xAI Grok Build CLI shows no verified benchmark rows. Google Gemini 3.5 Flash shows no verified benchmark rows. Google Gemini 3.1 Pro shows no verified benchmark rows. DeepSeek deepseek-v4-flash-0731 shows no verified benchmark rows. DeepSeek DeepSeek V4 Pro shows no verified benchmark rows. Qwen qwen3.7-flash shows no verified benchmark rows. Qwen qwen3.7-plus shows no verified benchmark rows. Qwen Qwen3.7 Max shows no verified benchmark rows. Moonshot kimi-k3 shows no verified benchmark rows. Moonshot Kimi K2.7 shows no verified benchmark rows. MiniMax M3 shows no verified benchmark rows.
Self-reported claims cover the entire dataset. Providers may publish internal numbers, yet no external audit validates those numbers. The absence of third-party sources leaves every numerical comparison unsupported. Anthropic entries carry zero verified benchmark rows. OpenAI entries carry zero verified benchmark rows. xAI entries carry zero verified benchmark rows. Google entries carry zero verified benchmark rows. DeepSeek entries carry zero verified benchmark rows. Qwen entries carry zero verified benchmark rows. Moonshot entries carry zero verified benchmark rows. MiniMax entries carry zero verified benchmark rows.
Researchers therefore restrict analysis to the explicit 2026-08-01 list. Origin balance, model count and CLI availability constitute the verifiable attributes. Any ranking derived from unverified claims violates the data constraints documented on the verification date.
Researchers track future third-party benchmark releases while testing models directly through available CLIs. The 2026-08-01 list contains 14 US models and 8 Chinese models. Direct CLI testing supplies the only current performance signal. Balance between origins remains the primary observable metric.
Track releases from independent labs that publish reproducible methodology. Monitor announcements from Anthropic, OpenAI, xAI, Google, DeepSeek, Alibaba, Moonshot and MiniMax for pricing updates. Subscribe to benchmark repositories that cover the listed model names.
Test coding tools in sequence. Install Cursor 2 and configure keys for claude-opus-5 and deepseek-v4-flash-0731. Run identical prompts and record output length and latency. Repeat the protocol with Claude Code for qwen3.7-plus and Grok Build CLI for grok-4.5. Install Windsurf and configure keys for Gemini 3.1 Pro and kimi-k3. Run identical prompts and record output length and latency. Repeat the protocol with Cline for Qwen3.7 Max and Aider for MiniMax M3. Aggregate results across at least three runs per model. Install GitHub Copilot and configure keys for GPT-5.3 Codex. Run identical prompts and record output length and latency. Install OpenAI Codex CLI and configure keys for GPT-5.3 Codex. Run identical prompts and record output length and latency. Install Gemini CLI and configure keys for Gemini 3.1 Pro. Run identical prompts and record output length and latency.
Compare origin balance by counting active models per provider. The current list shows equal emphasis on US and Chinese entries. Maintain this balance when selecting evaluation targets until verifiable metrics appear.
Additional reading appears in Fable 5 Benchmarks 2026: Ultimate Hands-On Analysis for AI Researchers and Ultimate Guide to Global Workspace Discovery in Claude: 2026 Research Implications for AI Tool Researchers. Researchers also consult Best AI Academic Writing Tools 2026: Ultimate Hands-On Review of Top Platforms for Research, Citation, and Scholarly Content Creation and Best AI Fact Checking Tools 2026: Ultimate Hands-On Review for Accurate Verification and Misinformation Detection for complementary evaluation frameworks.
The 2026-08-01 list shows multiple models from both US and Chinese labs with no verifiable performance metrics available, so no definitive lead can be determined.
No independently verified benchmarks exist for any models in the current frontier list; all potential numbers would be self-reported by the companies.
Focus on explicit model listings, origin balance, and direct testing through coding CLIs while awaiting future third-party evaluations.
Researchers can start with Cursor 2, Claude Code, Grok Build CLI, and similar cross-border options that support the listed models.
All pricing details remain unverified at this time; monitor official provider announcements for updates.
Explore more AI tools and guides
I spend $5,000+ monthly on AI subscriptions so you don’t have to. Every review comes from hands-on experience — not marketing claims.
Anthropic's potential global workspace advancements in Claude could reshape AI interpretability. This news analysis delivers unique angles for tool researchers evaluating frontier models in 2026. Discover actionable recommendations and future tool impacts.
A data-driven look at the 2026-07-01 frontier LLM list featuring Claude Fable 5 alongside top models. Discover evaluation strategies tailored for AI tool researchers and buyers seeking transparent comparisons.
In the fast-evolving world of academia, AI academic writing tools are revolutionizing how researchers draft papers, manage citations, and ensure scholarly integrity. This ultimate 2026 review hands-on tests the top platforms, highlighting features that streamline peer-reviewed workflows. Whether you're tackling literature reviews or thesis writing, find expert comparisons to elevate your productivity.
Join builders getting hands-on AI tool analysis — never sponsored, always tested.