Foundation Model Landscape

Alibaba’s Qwen3 family spanning Qwen3.8-Max (2.4T MoE / 95B active; live cloud alias qwen3.8-max routes to the 0902 snapshot as of 2026-09-05), open Qwen3.8-27B (dense VLM, Apache-2.0), and Qwen3.8-Flash-Next (125B / 6B active multimodal MoE + 51B n-gram embeddings)—a Qwen4 architecture preview for cost-efficient agentic coding. Production Qwen3.8-Flash on QwenCloud adds 1M-default context and built-in tools atop the Flash-Next design.

Anthropic’s Claude Opus 5 tier for complex agentic coding, enterprise work, long-context analysis, and careful instruction following. Claude Fable 5.1 sits above Opus for peak widely released capability.

Anthropic’s Claude Sonnet 5 tier — best combination of speed and intelligence for most production agents and coding, at lower cost than Opus.

Anthropic’s Claude Fable 5.1 — the most capable widely released Claude for long-horizon agents, deep reasoning, and demanding coding workflows. Mythos 5.1 is the same underlying model with relaxed cyber/life-sciences safeguards for trusted-access programs.

Anthropic’s fast, cost-efficient Claude tier for high-volume chat, classification, extraction, and sub-agent steps where latency and price matter more than peak reasoning.

DeepSeek’s reasoning-focused model trained with reinforcement learning for multi-step math, science, and coding problem solving.

DeepSeek’s V4 generation — live Flash is V4.1-Flash (API id deepseek-flash): Causal Encoder–Decoder MoE, 552B backbone (8B active prefill / 16B decode), native image+text, and 1M context. MIT weights on Hugging Face. Legacy deepseek-v4-flash and deepseek-v4-flash-vision-exp aliases temporarily route here. deepseek-v4-pro still serves V4-Pro-0813 after 2026-09-14 at unchanged Pro rates (DeepSeek withdrew the planned Flash reroute).

Google’s current Pro-class Gemini for hard reasoning and native multimodal work. Prefer API id gemini-3.1-pro-preview; Gemini 3.5 Pro remains partner-testing. Legacy gemini-2.5-pro is scheduled for shutdown Oct 16, 2026.

Meta’s Llama 4 family — open-weight multimodal models designed for research and commercial use under Meta’s community license.

Meta Superintelligence Labs’ Muse Spark 1.3 — closed multimodal reasoning model for agentic tasks, long-horizon coding, computer use, and 1M-context workflows via Muse Code and the Meta Model API. Succeeds Spark 1.2; max reasoning is still withheld for safety testing.

Meta Superintelligence Labs’ Muse Glimmer — Apache-2.0 ~30B dense multimodal agent model for on-device and single-GPU local agents. Sibling to closed Muse Spark; distinct from Llama 4.

Moonshot’s Kimi K3 — 2.8T MoE (104B active) open-weight multimodal agentic model with 1M context, native vision, and strong long-horizon coding. Weights on Hugging Face under the Kimi K3 License.

OpenAI’s GPT-5.6 family (Sol flagship, Terra balanced, Luna cost-efficient) for complex reasoning, coding, multimodal understanding, and agentic tool use. The gpt-5.6 API alias routes to Sol.

Meta’s Llama 4 family — open-weight multimodal models designed for research and commercial use under Meta’s community license.

DeepSeek’s reasoning-focused model trained with reinforcement learning for multi-step math, science, and coding problem solving.

DeepSeek’s V4 generation — live Flash is V4.1-Flash (API id deepseek-flash): Causal Encoder–Decoder MoE, 552B backbone (8B active prefill / 16B decode), native image+text, and 1M context. MIT weights on Hugging Face. Legacy deepseek-v4-flash and deepseek-v4-flash-vision-exp aliases temporarily route here. deepseek-v4-pro still serves V4-Pro-0813 after 2026-09-14 at unchanged Pro rates (DeepSeek withdrew the planned Flash reroute).

Moonshot’s Kimi K3 — 2.8T MoE (104B active) open-weight multimodal agentic model with 1M context, native vision, and strong long-horizon coding. Weights on Hugging Face under the Kimi K3 License.

Meta Superintelligence Labs’ Muse Glimmer — Apache-2.0 ~30B dense multimodal agent model for on-device and single-GPU local agents. Sibling to closed Muse Spark; distinct from Llama 4.

Alibaba’s Qwen3 family spanning Qwen3.8-Max (2.4T MoE / 95B active; live cloud alias qwen3.8-max routes to the 0902 snapshot as of 2026-09-05), open Qwen3.8-27B (dense VLM, Apache-2.0), and Qwen3.8-Flash-Next (125B / 6B active multimodal MoE + 51B n-gram embeddings)—a Qwen4 architecture preview for cost-efficient agentic coding. Production Qwen3.8-Flash on QwenCloud adds 1M-default context and built-in tools atop the Flash-Next design.

OpenAI’s GPT-5.6 family (Sol flagship, Terra balanced, Luna cost-efficient) for complex reasoning, coding, multimodal understanding, and agentic tool use. The gpt-5.6 API alias routes to Sol.

Anthropic’s Claude Opus 5 tier for complex agentic coding, enterprise work, long-context analysis, and careful instruction following. Claude Fable 5.1 sits above Opus for peak widely released capability.

Anthropic’s Claude Sonnet 5 tier — best combination of speed and intelligence for most production agents and coding, at lower cost than Opus.

Anthropic’s Claude Fable 5.1 — the most capable widely released Claude for long-horizon agents, deep reasoning, and demanding coding workflows. Mythos 5.1 is the same underlying model with relaxed cyber/life-sciences safeguards for trusted-access programs.

Anthropic’s fast, cost-efficient Claude tier for high-volume chat, classification, extraction, and sub-agent steps where latency and price matter more than peak reasoning.

Google’s current Pro-class Gemini for hard reasoning and native multimodal work. Prefer API id gemini-3.1-pro-preview; Gemini 3.5 Pro remains partner-testing. Legacy gemini-2.5-pro is scheduled for shutdown Oct 16, 2026.

Meta Superintelligence Labs’ Muse Spark 1.3 — closed multimodal reasoning model for agentic tasks, long-horizon coding, computer use, and 1M-context workflows via Muse Code and the Meta Model API. Succeeds Spark 1.2; max reasoning is still withheld for safety testing.